Proof Holdings

Create a new identity verification. Initiates a verification flow for an asset identifier (phone number, email, domain, social handle, wallet address, account, or Telegram bot). Agent usage: pass the asset `type`, the delivery `channel`, and the `identifier` (e.g. type "phone", channel "sms", identifier "+37060000000"; or type "email", channel "email", identifier "[email protected]"). The response carries a `challenge` object with the instruction for the chosen channel — except a `telegram_bot` verification made with a LIVE key, which the server completes on the spot and answers with `status: verified` plus a proof token and no challenge at all; with a `pk_test_` key that same call stays pending and does carry a challenge. On the messenger channels it also carries `challenge.deep_link` — pass THAT to render_auth_link to give the user a clickable link and a QR code. This endpoint returns no `qr_text` at all. For OTP-based flows (sms/email), the user receives a code — poll get_verification or use wait_for_verification to track completion. For Telegram/WhatsApp, the user clicks the deep link to verify. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a verification by ID. Returns the full verification object including status, channel, and proof token if verified. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List verifications with optional filters. Returns paginated results. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Trigger a DNS/HTTP verification check for a pending domain verification. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Submit an OTP/challenge code for a pending verification. Accepted ONLY where we delivered the code out-of-band to the party being verified: type "email", and type "domain" with channel "email". Any other pair answers 400 unsupported_for_type and completes elsewhere — use trigger_verification (POST /verifications/:id/verify) for a domain over dns/http/auto; a phone completes when the code reaches us FROM that number; social completes through its OAuth callback; a telegram_bot is already verified at creation. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Resend a verification message (email channel only). Generates and sends a new code. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Auto-complete a verification in test mode (pk_test_* API keys only). Useful for testing flows without real OTP delivery. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List verified users grouped by external_user_id. Shows all verifications per user. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a single verified user and their verifications by external user ID. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Start a B2B domain verification. Creates a verification record and returns DNS/HTTP instructions for the customer to complete. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Check the status of a pending domain verification. Triggers a DNS/HTTP check and returns the updated status. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete a domain verification attempt that is not verified (pending, verifying or failed) — nothing is revoked. A verified domain, or one that needs re-verification, is refused with domain_verified: removing it revokes its proofs and every delegation issued from it, which is delete_domain in a start_login session (or the dashboard). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Poll a verification until it reaches a terminal state (verified, failed, expired, revoked, or cancelled). Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. A multi-channel verification that lost the race is "cancelled" and one whose proof was withdrawn is "revoked"; both are final, so the wait returns rather than polling on. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a multi-channel phone verification: one phone number challenged over several channels at once (OR / first-wins). The first channel the user completes wins, yields a single proof token, and cancels the losing siblings. Exactly one proof is billed, at the winning channel's completion — not one per channel, and nothing at creation. Agent usage: the response includes a `group_id` (format vg_<hex>) and a `channels` array — each block carries a `deep_link` (WhatsApp/Telegram) or an `sms_message` to present to the user — pass the `deep_link` to render_auth_link. A block's `qr_text` is that same link already rendered as QR art, not an alternative address: print it verbatim inside a fenced code block or leave it out. Poll get_multi_channel_verification_status with the group_id to track completion and retrieve the proof token when a channel verifies. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the status of a multi-channel verification group by its group_id. Returns the aggregate status (pending/verified/expired), the winning_channel once one completes, per-channel statuses, and the proof token (with expires_at) when the group is verified. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a multi-asset verification request. Allows verifying multiple assets (phone, email, domain, social, wallet, account, telegram_bot) in a single flow. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a verification request by ID. Returns the request with all its asset verification statuses. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get proof tokens for verified assets in a verification request. Returns proof tokens that can be used for offline verification. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List verification requests with optional filters. Returns paginated results. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a verification request by its reference ID (your unique identifier). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Cancel a pending verification request. Only pending requests can be cancelled. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Poll a verification request until it reaches a terminal state (completed, expired, or cancelled). Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Validate a proof token online. Checks the token signature, expiry, and revocation status against the server, and optionally that the proof is about the identifier you expect. Public endpoint — no API key required.

Get the status of a proof by its public handle (ph_ctl_* / ph_dlg_*) or a verification ID. Returns whether the proof is active, suspended, expired, or revoked — a live proof reads "active". ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Revoke a proof by its public handle (ph_ctl_* / ph_dlg_*) or a verification ID — the same addresses get_proof_status takes. The proof token will no longer validate anywhere: the status endpoint, the public validator, the revocation list and the Token Status List all report it revoked. A domain CONTROL proof (a ph_ctl_* of a domain) also revokes, for good, every delegation issued from it; for a domain you own, every proof of the domain is revoked and the domain needs re-verification. The API key needs proofs:write, plus delegations:write while the revocation reaches live delegations — for a domain you own, any live delegation of the domain, whichever of its handles you name (the refusal names the scope and the count). A handle revoked before the domain was verified again does not reach the new proof: revoke that one by its own handle. Otherwise it revokes the PROOF only — the asset or consent it is about stays. Need a temporary stop? Use suspend_proof instead. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the revocation list for offline proof validation. Returns a signed list of all revoked proof IDs. Reuse a fetched list for up to 5 minutes (its cache lifetime), then fetch it again to see new revocations; if you cannot, its signature stays verifiable offline for 1 hour.

Takes no parameters.

Get the current token of a proof you own, by its public ph_ctl_* handle. After a signing-key rotation this is the re-signed token, which replaces any older copy you hold. Each read is audited. A revoked or expired proof answers proof_not_live; a suspended one is still served. The proof of one of the account's own verified assets (a domain, phone or email verified on the account itself) answers owner_session_required here whatever its status (with a live key; a test key gets not_found): this tool sends the API key, and such a token is read by the signed-in owner in the dashboard. A key bound to a sub-account reads only that sub-account's proofs; another's answers not_found. For a delegation (ph_dlg_*) use get_delegation_token or get_delegation instead. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List your proofs whose token was re-signed at or after an instant, oldest first — what to re-fetch after a signing-key rotation, for example when a webhook was missed. Items are references {proof_id, reissued_at, token_url}, never tokens: pull a ph_ctl_* item with get_proof_token and a ph_dlg_* item with get_delegation_token. Proofs of the account's own verified assets are not listed here (the API key cannot read their tokens); the dashboard shows them. A key bound to a sub-account lists only that sub-account's proofs and the account's delegations. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the current token of an active delegation by its public ph_dlg_* handle — after a signing-key rotation, the re-signed token to publish in place of the old one. Public endpoint — no API key required, since the token is already published with the delegation. Any handle that is not an active delegation (revoked, suspended, expired or unknown) answers the same not_found.

Create a new phone verification session. Returns deep_link, qr_code (base64 PNG), and qr_text (UTF-8 text QR for terminal display). Sessions provide a hosted verification flow with callbacks. Agent usage: After creating a session, pass the response's `deep_link` to render_auth_link so the user can open or scan it. Never hand `qr_text` to a link renderer — it is that link already rendered as QR art. Print it verbatim inside a fenced code block only when a real terminal needs the QR. On sms the person texts a short code from the number being proven, as in SMS login: `deep_link` is an `sms:` link to our default number with the message pre-filled, so render it the same way, and also show `sms_message` (what they send) with the numbers in `sms_dids` (one per region, each with its own QR), because a person reading on a computer texts from their phone. They get no reply SMS. An SMS proof can still be revoked about a minute after the session completes, if the delayed spoof check finds the sender was not a real mobile line, so before relying on it check it with validate_proof. Then use wait_for_session to poll until the session reaches a terminal state (verified, failed, or expired). The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a session by ID. Returns the session status, verification details, and proof token if verified. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Poll a session until it reaches a terminal state (verified, failed, or expired). Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List all verified assets for the authenticated user. Assets are verified identities (phones, emails, domains). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a single verified asset by ID. Returns details including type, value, status, and verification timestamps. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Revoke the proofs of one of your assets (a phone, email or domain) — the asset stays on the account. Owner only: sign in with start_login (a connection with only an API key is refused user_session_required — a key revokes a proof with revoke_proof instead), then one code: start_2fa_for_action with action_type revocation_bulk and targets [this asset id], wait_for_2fa, and call again. A phone or email: every proof of that number or address is revoked; signing in with it again issues a fresh proof. A domain you own: every proof of the domain and every delegation issued from it are revoked for good, and the domain needs re-verification — re-verifying does NOT bring the delegations back. Need a temporary stop for a domain proof? Use suspend_proof instead. Status-checking verifiers see the revocation within about a minute; signature-only verifiers accept a token until it expires. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Get the signed-in user's account details, plan and settings. Needs a login session: call start_login, then wait_for_login until it reports success. An API key alone cannot read this — for the key's own account use get_self_api_key or get_settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

List the current user's signed-in devices: every session that can still renew, idle ones included. Each shows its channel, user agent and expires_at, which is when the session stops being able to renew (not when its current access token expires). When called via API key, no session is marked is_current. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Sign out one device (session) by ID. Over a signed-in session it cannot end the current one (that is logout). Over an API key it can end any of the account's sessions. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get usage metrics for the authenticated account. Shows verification counts, API calls, and quota usage. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a one-call account snapshot: quota with end-of-month projection, request health (pending/completed/expired counts + success rate), per-channel completion rates, HITL counts, and active API key count. The same data the dashboard home renders. The hitl_channels sub-object requires the hitl:read scope. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get account settings: the outbound webhook endpoint with its event subscriptions, and the default proof lifetime. Branding (name, logo, colour, email theme, support email) belongs to each sub-account: read it with get_profile. The webhook signing secret is NOT returned to an API key — the response reports webhook.secret_set instead, and the value itself is visible only in the dashboard or in the update_settings response that created or rotated it. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Update account settings: the outbound webhook endpoint with its event subscriptions, and the default proof lifetime. Branding belongs to each sub-account: change it with update_profile. Setting the first webhook endpoint creates the signing secret and returns it in this response; the value is returned only when this call created or rotated it, so record it then. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Export a complete copy of your account's data (GDPR Articles 15 and 20), or check on one. Owner only: sign in with start_login first (a connection with only an API key is refused user_session_required). The first call starts the export; it is prepared in the background, and the account owner is told by email (or on their login channel) with a link that downloads it only while they are signed in — this tool never returns the data itself. Call again to see whether it is ready: an export already preparing or ready is reported instead of starting another. One new export per 24 hours (no limit while the account is scheduled for deletion); the file is kept for 7 days. Pass export_id to read one export's status. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Prefix-anchored global account search across six resource buckets (verification_requests, verifications, hitl, confirmations, authorizations, api_keys). Case-sensitive prefix match over indexed plaintext fields only; recipient identifiers are masked. Results are scoped to the authenticated account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the calling API key's own redacted metadata (id, name, key_prefix, environment, scopes, last_used_at, revoked_at, created_at) for agent self-inspection. Never returns the key hash or full secret. Requires API-key auth — a JWT session has no single-key context (returns 400 api_key_required). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Get per-API-key usage for a given key id: verification counts by type/channel/status, plus verification-request/confirmation/authorization totals, and attribution context (attribution_cutoff_at, attributed_fraction). The counts cover the key and every key it replaced (api_key.replaced_key_ids). Counts only — no recipient identifiers. Scoped to the authenticated account; an API-key caller may read only its own key. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List the custom message templates of one sub-account, organized by channel and message type. Templates belong to a sub-account, not to the account: profile_id picks it — omitted, a sub-account key lists its own sub-account and an account-wide key the primary sub-account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the default message templates — one for each channel and message type a message is actually sent for. These are the built-in templates a sub-account sends when it has no custom template of its own. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Get the template a sub-account sends for a channel and message type: its custom template if set (scope "project"), otherwise the default (scope "default"). The response names the sub-account in profile_id. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create or update a sub-account's custom template for a channel and message type, replacing the default for that sub-account only. Name the brand with {{business_name}}: a template spelling out a sub-account's own name is refused (template_hardcoded_name). A template for a channel and message type no message is sent for, or using a variable that message never fills, is still saved, with an entry in `warnings` (template_not_sent / template_variable_not_filled) saying why it will not be sent as written (an API that predates `warnings` omits the field). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete a sub-account's custom template, reverting that sub-account to the default template for the channel and message type. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Preview a template with sample data before saving, as a sub-account would send it. Shows how the message will look with placeholder variables filled in. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Render the template a sub-account sends — its custom template, otherwise the default — with provided variables. Returns the fully rendered message content. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get webhook delivery statistics including total deliveries, success/failure counts, and recent activity. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

List webhook delivery attempts with pagination and filtering. Shows delivery status, response codes, and timestamps. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get details of a single webhook delivery attempt including request/response payloads and retry history. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Retry a failed webhook delivery. Creates a new delivery attempt with the same payload to the configured endpoint. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List the account's webhook endpoints (URL, events, scope, sub-accounts, status). Secrets are never listed. A sub-account key sees only the endpoints scoped to exactly its own sub-account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Get one webhook endpoint by ID. The signing secret is not returned (only create_webhook_endpoint and rotate_webhook_endpoint_secret return it). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Add a webhook endpoint. An event reaches every active endpoint whose scope includes the event's sub-account and whose events match; an account-level event reaches "all"-scope endpoints only. An account still delivering to the single webhook URL in its settings is moved onto endpoints by this call (that URL becomes an "all"-scope endpoint with the same secret and events). The response carries the endpoint's signing secret — this and rotate_webhook_endpoint_secret are the only times it is returned, so store it now. Deliveries carry an HMAC-SHA256 X-Proof-Signature over "<X-Proof-Timestamp>.<body>" made with it. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Update a webhook endpoint: its URL, events, scope, sub-accounts, status or description. Every field is optional, but send at least one. Changing the scope to "selected" without profile_ids keeps the stored list. The signing secret is unchanged (rotate it with rotate_webhook_endpoint_secret). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete a webhook endpoint. It receives no new events; deliveries already queued to it are still attempted, signed with the secret they were queued with. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Replace a webhook endpoint's signing secret and return the new one — this and create_webhook_endpoint are the only times a secret is returned, so store it now. New deliveries are signed with the new secret; deliveries already queued keep the secret they were queued with. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the current subscription details including plan, status, billing period, and usage limits. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Get the prepaid balance (in euro cents and mills), the low-balance threshold, the top-up amounts on offer and the auto top-up setting. Proofs past the plan's included amount are drawn from this balance; once billing is live, an account whose balance is at or below zero is refused new proof requests (402 balance_empty) until it tops up. billing_enabled says whether billing is live. Topping up and changing auto top-up are done by the owner in the dashboard. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

List the prepaid balance history, newest first: every top-up, refund and adjustment, and one usage row per month for that month's deductions. Amounts are signed euro cents (credits positive). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List all profiles for the authenticated account. Returns all profiles including primary and secondary. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Create a new profile with optional display name, bio, avatar, business info, email branding (theme, support email), and theme settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a specific profile by ID. Returns full profile details including theme, custom links, and verification display settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Update a profile by ID. Supports changing display info, email branding (theme, support email), theme, custom links, and verification display settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete a sub-account (profile). Revokes its API keys, archives the HITL setups and declines the pending requests created with its own keys, keeps its history but unlinks it, deletes its message templates, and retires its username for good. The last sub-account cannot be deleted; deleting the primary sub-account makes the oldest remaining one primary (the response names it). In a start_login session: one code first — start_2fa_for_action with action_type profile_delete and targets [this sub-account id]. With an API key alone only an EMPTY sub-account (no API keys, HITL setups or pending requests) can be deleted; anything else answers owner_session_required. archive_default_linked is deliberately not offered here: archiving or cancelling everything linked only by default is an irreversible fan-out, so archive each item explicitly instead. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Set a profile as the primary profile. The previous primary profile becomes secondary. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Update which verified assets (proofs) are displayed on a specific profile. Controls visibility, privacy masking, ordering, and labels. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the current user's primary profile. Returns profile details, theme, custom links, and public proof display settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Update the current user's primary profile. Supports display name, bio, avatar, theme, custom links, and verification display settings. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get available verified assets that can be displayed on the user's profile. Returns assets eligible for public proof display. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Rotate a sub-account's callback signing secret — the secret that signs every callback_url delivery of its verification requests — and return the new one. This is the only time it is returned, so store it now. Callback deliveries already queued keep the secret they were signed with. An account still delivering through its settings webhook URL is switched over to webhook endpoints first. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List all email addresses associated with the authenticated account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Remove an email address from the account. Always requires one code: start_2fa_for_action with action_type email_remove and targets [this email id]. Revokes every proof of the address and your own consents naming it, declines pending confirmations waiting on it, moves Lookalike Watch alerts to your remaining primary email (or stops the watch when none is left), deletes the record, signs out your other devices, and names the new primary email. Refused for the address you signed in with and for your only way to sign in. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Set an email address as the primary email for the account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Start adding a new email address. Sends an OTP code to the email. Returns a session ID to verify with. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Check the status of an in-progress email addition by session ID. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Submit the OTP code to verify an email addition. Completes the add-email flow if the code is correct. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Resend the OTP code for an in-progress email addition. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List all phone numbers associated with the authenticated account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Remove a phone number from the account. Always requires one code: start_2fa_for_action with action_type phone_remove and targets [this phone id]. Revokes every proof of the number and your own consents naming it, declines pending confirmations waiting on it, flags approval setups left with no approver, deletes the record and its messenger links, signs out your other devices, and names the new primary phone — your Proof of Me circles then reach you there. Refused for the number you signed in with, for your only way to sign in, and for your last phone while circles are active (circles_need_phone: add another phone first, or archive the circles). Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Set a phone number as the primary phone for the account. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Start adding a new phone number by reverse OTP: the user starts the exchange themselves and nothing is sent before they do; the response's `instructions` field says what they do (on telegram the bot then asks them to tap a button that shares their number). Returns a session ID; on telegram/whatsapp also deep_link, qr_code (base64 PNG) and qr_text (UTF-8 text QR), and on sms sms_message and sms_dids (one entry per region; a region with no number configured has an empty `did`) with no deep_link. Agent usage: on telegram/whatsapp pass `deep_link` to render_auth_link for the link and a QR code. On sms there is no deep_link — show `sms_message` and a number from an `sms_dids` entry whose `did` is non-empty. Never hand a `qr_text` to a link renderer — it is a link already rendered as QR art. Then poll get_add_phone_status with the session ID. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Check the status of an in-progress phone addition by session ID. Poll until terminal state. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List all API keys for the authenticated account. Returns key metadata (not the secret key values). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Create a new API key. The secret key value is returned ONLY in this response — store it securely. Note: the secret will be visible in the AI conversation context. Optionally bind the key to a sub-account via profile_id — outbound messages sent with the key are then branded as that sub-account. Agent usage: This operation requires 2FA. Before calling, complete the 2FA flow: (1) call start_2fa with action_type "api_key_create" and the user's preferred channel, (2) tell the user a code was sent and wait for them to verify, (3) poll get_2fa_status until status is "verified", (4) then call create_api_key — the backend session will be elevated. If you get a 403, the 2FA session has expired — restart the flow. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Revoke an API key by ID. The key will immediately stop working. This action cannot be undone. Agent usage: This operation requires 2FA. Before calling, complete the 2FA flow: (1) call start_2fa with action_type "api_key_revoke", (2) wait for user to verify the code, (3) poll get_2fa_status until "verified", (4) then call revoke_api_key. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Regenerate an API key: issues a new key (a new id and secret) with the same name, environment, scopes and sub-account. The old key stops working immediately; use the id in this response from now on. The new key lists the old one in replaced_key_ids, so its requests and usage stay with it. The new secret is returned ONLY in this response. Note: the secret will be visible in the AI conversation context. Agent usage: This operation requires 2FA. Before calling, complete the 2FA flow: (1) call start_2fa with action_type "api_key_regenerate", (2) wait for user to verify the code, (3) poll get_2fa_status until "verified", (4) then call regenerate_api_key. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Start a two-factor authentication challenge for a sensitive operation. Sends a verification code via the chosen channel. Returns a session ID, and for messaging channels: deep_link, qr_code (base64 PNG), and qr_text (UTF-8 text QR for terminal display). Agent usage — full 2FA flow: (1) Call start_2fa with the appropriate action_type and channel. (2) Present the challenge, and do NOT announce a message the server did not send: email is the ONLY channel it dispatches on. On telegram/whatsapp the user opens `deep_link` and sends the message themselves; on sms they send `sms_message` to a number from `sms_dids`; the response's own `instructions` field says which. On email, a 200 carrying `requires_email_selection: true` and `available_emails` means nothing was sent and no session exists — ask which address and call again with `email_id`. For telegram/whatsapp, pass `deep_link` to render_auth_link; for sms, show `sms_message`. Never hand `qr_text` to a link renderer — it is the link already rendered as QR art, so print it verbatim inside a fenced code block only when a real terminal needs the QR. (3) Poll get_2fa_status with the returned session_id until status is "verified" (the user enters the code on their device) or "expired". (4) If verified, proceed with the protected operation (e.g. create_api_key). If expired, inform the user and offer to restart. Typical channels: "telegram" or "email". For email, the user may receive a magic link instead of a code — the backend handles this automatically. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Check the status of a 2FA session by session ID. Returns whether the challenge has been completed, is pending, or has expired. Agent usage: Poll this after calling start_2fa. Check every few seconds. Terminal states: "verified" (proceed with protected operation), "expired" (restart with start_2fa). While "pending", the user has not yet entered their code. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Submit a verification code to complete a 2FA challenge. Rate limited to 10 attempts per minute. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Complete a 2FA challenge using a magic link token (received via email). No code submission needed — the token itself is the proof. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Poll a 2FA session until the user completes the challenge. Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Agent usage: Call this after start_2fa with the returned session_id. Terminal states: "verified" (challenge completed — proceed with the protected operation), "expired" (offer to restart with start_2fa). Before polling, present start_2fa's challenge: pass its `deep_link` to render_auth_link on telegram/whatsapp, or show its `sms_message` on sms. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List all domains for the authenticated account. Returns domain metadata including verification status. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Add a new domain to verify ownership. Optionally specify email sending intent and verification method. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Get details of a specific domain by ID, including verification status, DNS records, and provider connections. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Remove a domain from the account. Revokes every proof of the domain and every delegation issued from it (for good — re-adding the domain does not bring them back), deletes the domain and its records, and removes our _proof TXT record and the Resend sending domain where we can; the response lists what is left for you to remove by hand. Owner only, in a start_login session, with one code: start_2fa_for_action with action_type domain_remove and targets [this domain id]. Need only a fresh proof? Revoke the proofs instead (revoke_asset), which keeps the domain. Cannot be undone. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Trigger domain verification. Checks DNS records or other verification method to confirm domain ownership. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Connect a Cloudflare API token to a domain for automated DNS record management. Note: the API token will be visible in the AI conversation context. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Connect GoDaddy API credentials to a domain for automated DNS record management. Note: the API key and secret will be visible in the AI conversation context. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Connect a generic DNS provider to a domain using provider-specific credentials. Note: credentials will be visible in the AI conversation context. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Add an additional DNS provider for verification to an already-verified domain. Note: credentials will be visible in the AI conversation context. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Get metadata for all supported DNS providers, including required credential fields and documentation links. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Verify a pending domain using previously stored DNS credentials. Use when credentials are already connected. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Check if stored DNS credentials have access to a domain. Use before triggering verification to confirm the credentials work. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Start domain verification via corporate email. Sends a verification code to a standard admin email address at the domain. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Confirm domain email verification by submitting the code sent to the corporate email address. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Resend the domain verification email. Rate limited to 3 requests per minute. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Register a verified domain with Proof's email provider (Resend) for sending. The result carries the DKIM/SPF DNS records to publish (`resend.dkim_records`) and their status. `from_email`, if given, must be on this domain and is stored as the preferred sender. Proof does not send from customer domains yet: verification emails still come from Proof's own address, and the stored `from_email` is not used. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Check the email sending setup status for a domain, including DNS record configuration progress. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List all stored DNS provider credentials for the authenticated account. Returns credential metadata (not secret values). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Store DNS provider credentials for automated domain verification. Note: credentials will be visible in the AI conversation context. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Delete a stored DNS provider credential. Domains using this credential will need new credentials for automated verification. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List verification requests created by the authenticated user. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

List incoming verification requests where the authenticated user is the subject. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Takes no parameters.

Create a new verification request to ask another user to share verified assets. Each asset is an object naming its type, e.g. { "type": "email" }. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Cancel a pending verification request. Only the creator can cancel. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Extend the expiration time of a pending verification request. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Email the link of a pending verification request to a recipient. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Start a self-service domain verification session. The user proves ownership of a domain via DNS or HTTP challenge. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Get the current status of a domain verification session, including challenge details. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Trigger a check on the domain verification challenge. Call after placing DNS record or HTTP file. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Claim a unique username for the authenticated user's public profile. Usernames must be 3-30 characters, alphanumeric and underscores only. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Update which verified proofs are visible on the user's public profile, including visibility, masking, and display order. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), or sign in with start_login — a session opens this tool — then call this tool again.

Create a new authorization request. Authorizations grant permission to send confirmations to a user via a specific channel. The user must approve the authorization before confirmations can be sent. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get an authorization by ID. Returns the current state of the authorization including status and usage statistics. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List authorizations with optional filters. Returns paginated results. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Revoke an active or pending authorization. Only authorizations with status "active" or "pending" can be revoked. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Export authorizations as CSV or JSON. Supports date range and status filtering. Maximum 10,000 records. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a HITL confirmation request. Sends an approval request to the channels configured on the referenced HITL config. First response wins. Agent usage: The HITL config referenced by hitl_id must be authorized before you can create confirmations. If this returns a 403 or "not authorized" error, call request_hitl_authorization on the HITL config first, then wait for the user to approve via the messaging channel. After creating a confirmation, poll get_confirmation to check for approval/denial, or, if the user (the HITL config owner) will decide it themselves, use get_confirmation_approval_link for their own approval link. Encryption is NOT optional: a HITL config cannot send confirmations without a keypair, the human approves by decrypting in their browser, and `message` must be a JSON ciphertext envelope. Plaintext is refused with 400 `message_not_encrypted`; an envelope sealed to another key with 400 `message_key_mismatch`. Choose the version by who must be able to decrypt: - v1 (single-recipient): wraps the content key for the config owner only. Use ONLY when there are no enrolled non-owner approvers — with approvers enrolled it is refused with 400 `envelope_missing_approvers`. - v2 (multi-recipient): additionally wraps the content key for each enrolled approver via an `approver_ek` map, so every approver can decrypt. Build it with the SDK `encryptMessageV2(plaintext, ownerPublicKey, hitlId, approverKeys)` after fetching the approvers via the SDK `getApproverKeys(hitlId)` (`GET /api/v1/hitl/:id/approver-keys`). AAD is the `hitlId`. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a confirmation by ID. Returns the full confirmation including status, response, and proof token. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List confirmations with optional filters. Returns paginated results with status, HITL config and sub-account (profile_id) filtering. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the approval link of the HITL config owner: it is decided with the config's own key, so it is for the owner to approve or deny the confirmation themselves. An approver cannot use it — an approver's link goes only to their own channel, and "Send again" (`POST /api/v1/confirmations/{id}/resend`, which no tool calls yet) is what sends it there again. Every call returns the same owner link until "Send again" reaches the owner; "Send again" replaces it, and the next call returns a new one. Calling this tool again cancels nothing, so it is safe to call again to show the link again. The link expires when the confirmation does. Only works for pending confirmations. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Poll a confirmation until the approver responds. Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Agent usage: Call this after create_confirmation with the returned confirmation ID. Terminal states: "approved" (action authorized — check the proof_token), "denied" (action rejected). While "pending", the approver has not yet responded. This poll returns no link — the approval request is delivered to the approver's own channel, and only the owner's "Send again" (`POST /api/v1/confirmations/{id}/resend`, which no tool calls yet) delivers it there again. If the owner will decide instead, call get_confirmation_approval_link for the owner's `approval_url`; it needs the confirmation to be still pending — the state this tool waits in — and also refuses once `timeout_at` has passed, so a long wait can outlive the link even while the status has not moved. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a HITL (Human-in-the-Loop) config. Defines which messaging channels receive approval requests and the default timeout. Agent usage: After creating a HITL config, you must call request_hitl_authorization before it can be used with create_confirmation. For Telegram channels, use create_chat_id_discovery first to discover the user's chat ID, then include it in the channel config. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a HITL config by ID. Returns the config including channels, timeout, and status. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List HITL configs with optional filters. Returns paginated results. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Update a HITL config. Can change name, channels, or timeout. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete (archive) a HITL config. Sets status to archived. Cannot be used for new confirmations after deletion. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Request authorization for a HITL config. Sends consent requests to all configured channels (Telegram/WhatsApp). Users must approve before confirmations can be sent. Agent usage: This must be completed before create_confirmation will work for this HITL config. The consent request is DELIVERED to each channel — the response carries no link to show. It returns `authorizations` — one entry per resolved recipient, each with its `channel`, `recipient`, `authorization_id` and `status` — plus a `message`. Read the entries rather than assuming: `status: 'pending'` means consent was just sent and the user must approve in Telegram/WhatsApp itself, while `status: 'active'` means that recipient had already authorized and nothing was sent to them. An EMPTY array does not mean everyone is authorized — it means no recipient could be resolved at all (usually a HITL config bound to a missing or inactive circle, or channels whose recipient is blank), so nobody was asked and waiting for an approval would hang forever. Do not relay `message` on that branch: it reads 'All channels already have active or pending authorizations', which is exactly the wrong conclusion. Read the array. After requesting, you can proceed once the user confirms they have approved. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a Telegram chat ID discovery token. Returns a deep_link, qr_code (base64 PNG), and qr_text (UTF-8 text QR for terminal display). The user opens the link in Telegram, which reveals their chat ID for HITL config setup. Agent usage: pass the response's `deep_link` to render_auth_link so the user can click or scan it. Never hand `qr_text` to a link renderer — it is that same link already rendered as QR art. Print it verbatim inside a fenced code block only when a real terminal needs the QR. Then poll poll_chat_id_discovery with the returned token until the chat_id appears. Use the discovered chat_id when creating a HITL config with a Telegram channel. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Takes no parameters.

Poll a chat ID discovery token. Returns the discovered Telegram chat ID, user ID, and username once the user interacts with the bot. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Poll a chat ID discovery token until the user interacts with the Telegram bot. Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Agent usage: Call this after create_chat_id_discovery. Present the discovery response's `deep_link` with render_auth_link first, then poll with the returned token. Terminal states: "completed" (chat_id discovered — use it when creating a HITL config with a Telegram channel), "expired". While "pending", the user has not yet opened the Telegram link. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get the encryption keypair for a HITL config. With `hitl:read` it returns the public key, encrypted private key, KDF salt, and key ID (kid). A key holding `confirmations:create` without `hitl:read` receives only `public_key` and `kid` — what a confirmation message is sealed to, never the private-key material. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Upload an encryption keypair for a HITL config. The server validates the RSA-4096 public key and computes the key ID. Note: the encrypted_private_key is already encrypted client-side with the user's password — the server never sees the password. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Delete the encryption keypair from a HITL config. WARNING: Existing ciphertexts will become permanently unreadable. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a Proof-Me cross-channel identity challenge (CONFIRM) for an enrolled Circle member. The account holder approves/denies on a channel different from the one the suspicious contact reached the member on. Agent usage: poll get_identity_challenge with the returned ID until it leaves pending (confirmed, denied, cancelled, failed, expired or revoked). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a Proof-Me identity challenge by ID. Returns its status (pending, confirmed, denied, cancelled, failed, expired, revoked or suspended), the channel the CONFIRM ran on, and — once the person being checked decided — the proof token and its proof_id. Poll this for resolution. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a Circle — a named group of trusted contacts (the people-primitive powering Proof-Me and Approvals-on-Circle). Optionally seed initial members; each member's declared channels are stored so they can be invited to enroll. Agent usage: after creating a Circle, add members with add_circle_member (or seed them here), then invite_circle_member to mint the enrollment deep link each contact taps to bind their messenger identity. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List Circles with optional filters. Archived Circles are excluded by default; pass status="archived" to surface them. Returns paginated results. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get a Circle by ID. Returns the Circle including its members and per-member channel presence booleans (an identifier is NEVER returned — SEC-PM-006). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Update a Circle's name and/or owner name. Omitted fields are left unchanged; the profile/identity binding is create-only. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Archive a Circle (soft delete — sets status to archived). It stops appearing in the default list but is recoverable via list_circles with status="archived". ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Add a pending member to a Circle by name, with an optional pre-declared WhatsApp phone. The Telegram chat id is captured later when the member taps the enrollment deep link. Agent usage: after adding a member, call invite_circle_member to mint the single-use enrollment link the member taps to bind their messenger identity. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List a Circle's members. Each member carries channel-presence booleans (has_telegram / has_whatsapp) and the masked phone they enrolled from (enrolled_phone_hint, e.g. +9••••••4171) — a full identifier is NEVER returned (SEC-PM-006). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Mint a single-use enrollment invite for a Circle member. Returns Telegram + WhatsApp deep links and a QR code; the member taps the channel-matched link to bind their messenger identity to the Circle. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Fire an on-demand Proof-Me drill against one enrolled member — a real cross-channel identity challenge carrying the scheduled-safety-check label (the account owner practices confirming). Complements the automated daily drill + monthly reinforcement sweeps. On Pay as you go a live key is refused `422 plan_upgrade_required`; a test key is not, and its drill reaches the member on Telegram only (`400 no_channel` when they have no verified Telegram channel). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Remove a member from a Circle. Deletes the member's channels and revokes any live authorizations that routed through them. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Declare a channel for a Circle member. Returns 409 if that channel type already exists on the member. The identifier is a Telegram chat id or an E.164 WhatsApp phone (per-channel format validated server-side). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List a Circle member's channels. Returns channel metadata only (channel, status, verified_at) — an identifier is NEVER returned (SEC-PM-006). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Remove a channel from a Circle member. Removing the last channel is allowed (the member simply becomes unreachable). ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Create a delegation (Proof of Delegation): a control-proven domain authorizes a typed url/purl artifact and receives a publishable signed token. The attestation is that the domain's controller authorized this artifact — nothing more; it says nothing about the artifact or the business behind it. Each mint is a billable proof. A requested lifetime longer than the control proof's remainder is rejected, never truncated. The result carries the publishable token AND the DERIVED effective_status/is_valid — the same pair the list and get tools return, so the shape does not depend on which verb produced it. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

List your own delegations with optional filters. Each item carries the DERIVED effective_status/is_valid beside its stored status — read effective_status, since status records only whether the owner revoked the delegation themselves, so a delegation still reads active there after its domain control proof was revoked, suspended or expired, and after its OWN lifetime ran out. effective_status accounts for all of those: it reports expired once either lifetime has passed, whichever is earlier. Tokens are not included in the list — use get_delegation for the stored token. There is no public enumeration of third-party delegations. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Get one of your delegations by Mongo id or ph_dlg_* handle. Returns the stored publishable token plus the DERIVED effective_status/is_valid — the delegation is only as live as the control proof it stands on (a revoked, suspended or expired domain proof invalidates it) and never outlives its own expires_at, whichever comes first. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Revoke ONE delegation. The domain control proof and sibling delegations stay valid; the revocation is mirrored into the proof registry so the CRL, status list and public status lookup all see it. Idempotent: a repeat revoke answers success with the same body, both here and through revoke_proof with the ph_dlg_ handle — unlike a PROOF, where a repeat answers already_revoked. The result carries the DERIVED effective_status/is_valid beside the stored status. ACCESS: needs a Proof account. Authenticate this client (Claude Code: /mcp → Authenticate), then call this tool again. start_login does NOT open this tool.

Verify a Proof of Delegation — the attestation that a domain authorized a specific agent artifact. Pass either the artifact's card (MCP server.json / A2A agent card) or a raw delegation token, PLUS two facts you established yourself: the artifact identity you resolved (the package you are installing, the endpoint you are calling) and the domain you expect to stand behind it. Both are required — a token can be copied into someone else's card, and any domain owner can mint a valid delegation naming someone else's package, so a verdict without both pins would mean 'some domain said something about some artifact'. WHERE THE IDENTITY MAY COME FROM: it is the artifact you are acting on, and never a value read from inside the artifact you are checking. A name in the artifact's own package.json, card or manifest is self-declared and editable by whoever ships it, so checking it against the delegation compares the artifact with itself. Resolve it afresh at call time instead of reusing a value from earlier in this conversation, which may already be stale. When no independently resolvable identity exists — a local or unpublished artifact — you have nothing to compare against, and that is the honest answer: report it rather than a mismatch, because a mismatch here reads as an accusation against the domain named in the delegation. The verifier walks signature → principal → delegate → revocation status, including the cascade down to the domain control proof. Requires no API key. What a verified result means: the expected domain's controller authorized this artifact — NOT that the artifact is safe, audited or endorsed.

Batch-verify up to 50 Proof of Delegation artifacts in one call — the inventory half of the checker (h-proof-checker-mcp): re-check everything you already trust instead of one verify_delegation call per artifact. Each item takes the same shape verify_delegation requires (card XOR token, delegate, expected_principal) and follows the same rule about where identity comes from: each `delegate` is the artifact you are acting on, never a value read from inside the artifact you are checking, and resolved afresh at call time. Items are verified INDEPENDENTLY — no cross-item state, nothing persisted, one item failing never affects another item's result. This tool does NOT discover what you have installed and does NOT fetch anything on your behalf: it verifies material you already hold, the same trust boundary verify_delegation draws. Revocation is always checked (there is no check_status:false here — the point of a batch re-check is to see what changed since you last looked). Each result carries an `outcome`: confirmed_valid (verified), confirmed_invalid (a genuine negative verdict — revoked, expired, mismatched, malformed, and similar), no_claim_found (the artifact publishes no delegation at all — an absence, never an accusation), or unconfirmed (we could not reach the issuer or otherwise get a confident answer right now — NEVER treat this as a bad verdict). Requires no API key. `checked_at` on each result is the moment THAT item's own check completed, not one timestamp shared across the call.

Start a login session by sending an authentication challenge to the user's chosen channel (Telegram, WhatsApp, SMS, or email). Returns a session ID and, FOR TELEGRAM AND WHATSAPP ONLY, deep_link, qr_code (base64 PNG) and qr_text (UTF-8 text QR for terminal display); on sms it returns sms_message with sms_dids instead, and on email nothing to display. Agent usage: (1) Call start_login with the desired channel and phone_number (for SMS) or email (for email). (2) Present the challenge, and WHICH FIELD depends on the channel. On telegram/whatsapp pass `deep_link` to render_auth_link, which returns the clickable link and a QR code to you; in a chat client the link is what the user acts on, since they are usually on the same machine. On sms `deep_link` is an EMPTY STRING and render_auth_link will reject it — show `sms_message` and let the user pick a number from `sms_dids`. `suggested_region` is always set, but the matching ENTRY in `sms_dids` may be missing (europe and israel appear only when a number is configured) or present with an empty `did`, so offer that region first only when `sms_dids[suggested_region]` exists and carries a number, and otherwise offer whatever the object does. On email there is nothing to display at all, and nothing to check either: a result means the mail was accepted for delivery, so tell the user to open their inbox. A delivery failure is an ERROR here, not a field — the tool answers `challenge_delivery_failed` and NO session exists, so do not call wait_for_login; retry, or offer another channel. Never hand `qr_text` to a link renderer — it is the link already rendered as QR art, so print it verbatim inside a fenced code block or not at all, because its rows stop scanning the moment one wraps or a blank line lands between them. (3) Once your reply shows the user the challenge, call wait_for_login with the returned session ID to poll until the user completes authentication. On email, tell the user to open the link and press "Sign in the brows

Poll a login session until the user completes authentication. Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Agent usage: Call this after start_login with the returned session ID. Terminal states: "verified" (login succeeded — this server keeps the session for this connection and signs your own account records with it: emails, phones, domains, API keys, 2FA, DNS credentials, verification requests, your public profile's username and proof list), "failed", "expired", "completed_elsewhere" (email only: the user pressed "Sign in here" on the email link's page, so the browser they clicked from is signed in and this connection never will be — tell them so, and start a new login if this connection still needs one). A result with status "pending" means the wait ran out before the user finished — call this tool again with the same id. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw.

Create an account-bootstrap session for agent-driven onboarding. Public endpoint (no API key required). Returns a session id, channel-specific instructions, and — FOR TELEGRAM AND WHATSAPP ONLY — deep_link, qr_code (base64 PNG) and qr_text (UTF-8 text QR). The endpoint is safe to call regardless of whether the submitted identity is new: completing verification transparently creates a User when none exists, or logs the user in when one already does. The response is intentionally uniform — the server does NOT reveal whether the identity is already registered (this would be a user-enumeration oracle). If the agent needs to distinguish new-vs-returning users, call `get_current_user` after verification and inspect the user's `created_at`. EMAIL CHANNEL SENDS NOTHING HERE. The response carries email_sent=false and send_required=true: confirm with the user that the address is theirs, then call send_account_email. The other channels dispatch nothing either way — the user sends the first message themselves (reverse OTP). Agent usage: (1) Call create_account with the user's chosen channel. (2) For email, confirm with the user and call send_account_email. On telegram/whatsapp pass `deep_link` to render_auth_link, which returns the clickable link with a QR code beneath it to you. On sms `deep_link` is an EMPTY STRING and render_auth_link will reject it — show `sms_message`, which is what the user sends; this response carries no destination number, so do not invent one. Do not hand `qr_text` to a link renderer either — it is the link already rendered as QR art; print it verbatim inside a fenced code block if a real terminal needs the QR. (3) Once your reply shows the user the challenge, call wait_for_account_creation with the session id until verification completes. On email the user must press "Sign in the browser that requested this" on the link's page — "Sign in here" signs in the browser they clicked from, and the wait then ends with "completed_elsewhere". (4) On verified

Deliver the verification email for an account-bootstrap session created with channel "email". Public endpoint (no API key required). Why this is a separate call: create_account deliberately sends NOTHING. A message from proof.holdings arriving in someone's inbox is an act with a person on the other end, so it takes an explicit second step rather than being a side effect of creating a session. ASK THE USER to confirm the address is theirs before calling this — do not call it automatically after create_account. The other channels (telegram, whatsapp, sms) never need this: the user sends the first message themselves (reverse OTP), so nothing is dispatched to them at all. One session sends at most one email, including under concurrent calls. A repeat answers 410 already_sent, which means the material backing the send is gone: usually because the email went out, occasionally because an attempt was interrupted mid-send and nothing was delivered. Do not tell the user a message was sent on the strength of a 410 — create a new session instead. (A session the recipient cancelled via the decline link, or an expired one, answers 404, not 410.) If delivery fails the response carries email_sent=false and the call can be retried.

Poll an account-bootstrap session until the user completes verification. Uses exponential backoff (2s initial, 1.5x, 30s max) with jitter. Agent usage: Call this after create_account with the returned session id. Terminal states: "verified" (this server keeps the session for this connection; the `user_id` field in the terminal body identifies the authenticated user), "expired", "completed_elsewhere" (email only: the user pressed "Sign in here" on the email link's page, so the browser they clicked from is signed in and this connection never will be — tell them so, and start a new login if this connection still needs one). A result with status "pending" means the wait ran out before the user finished — call this tool again with the same id. Works identically whether the underlying flow created a new User or logged in an existing one. Before calling this for a challenge you created, make sure your reply text already shows the user what to act on — the link itself, the SMS text and number, or a note to check their inbox; a link that appeared only in a tool result is one they never saw.

Start a 2FA challenge that grants the authenticated user permission to perform a sensitive action. The 2FA must complete on a channel DIFFERENT from the user's last login channel (the backend enforces this). On success a grant is stored keyed by the user id and action_type; middleware on the protected endpoint reads the grant and allows the mutation. Destructive action types (api_key_create, api_key_revoke, api_key_regenerate, api_key_view, account_delete, phone_remove, email_remove, phone_primary, email_primary, revocation_bulk, domain_remove, profile_delete) are capped at 5min TTL and forced single-use by server policy. A step-up for revocation_bulk, phone_remove, email_remove, domain_remove, profile_delete or account_delete is bound to `targets` — the exact ids the protected call will carry. Configuration scopes (settings_write, profile_update, templates_write) allow up to 24h TTL and caller-chosen single_use — useful for agents making multiple settings mutations in one session. Agent usage: (1) Attempt the sensitive call; if you get `403 2fa_required` note the `action_type`. (2) Call start_2fa_for_action with that action_type and a channel different from the login channel. (3) Present the challenge: on telegram/whatsapp pass `deep_link` to render_auth_link; on sms show `sms_message` and the number to send it to; on email check the response first: with more than one verified email and no `email_id` given, it answers 200 with `requires_email_selection: true` and `available_emails` — NO session was started and nothing was sent, so ask the user which address and call again with `email_id` rather than telling them to open a mailbox. Never hand `qr_text` to a link renderer — it is the link already rendered as QR art. Print it verbatim inside a fenced code block only when a real terminal needs the QR. (4) Call wait_for_2fa until the session is verified. (5) Retry the original sensitive call. The user does not see tool results — most clients fold them away, render_au

Render an authorization or verification URL as a clickable link followed by a scannable QR code inside a fenced code block. This is a local computation tool — no HTTP request is made. Its output comes back to you, not to the user: copy the link, and the QR block inside its fence, into your reply. Agent usage: pass a LINK field from an earlier response — `deep_link` (create_session, create_chat_id_discovery always; start_login, create_account, start_2fa, start_2fa_for_action, start_add_phone on telegram/whatsapp ONLY, where on sms and email that field is empty or absent altogether and this tool rejects it either way), `challenge.deep_link` (create_verification, messenger channels only), or `telegram_deep_link` / `whatsapp_deep_link`. Do NOT pass `qr_text` to this tool: that field is a QR code already rendered as text, and this tool takes a link. When a response gives you `qr_text` and no link, print it verbatim inside a fenced code block instead — its rows only scan while they stay adjacent, so a blank line or wrapped row destroys the code. The user does not see tool results — most clients fold them away, render_auth_link's output included. Before you wait or poll, put what the user must act on into your own reply text: the link itself (not a sentence announcing it), the SMS text and number, or a note to check their inbox.