SocialSeal

Read a saved operation's status, price, retained result and USD receipt. Polling and retrieving prior results are free. A funding_required operation has not started: after the workspace is topped up, send the same request again with the same idempotency key. Funding alone is not successful task completion.

Read the workspace's prepaid USD balance, the amount available for new actions (the balance minus the prices of admitted actions still running) and this month's spend, plus any one-time starter grant (starterGrantUsd), which only some eligible Free owners receive, and complimentaryVisibilityCheck (whether this user can still start the free Visibility Check). Read it before promising starter usage or that an action fits within the grant. Subscription credits remain their own allowance and are never converted to a cash balance.

Receive one complete evidence-backed Visibility Check per eligible user for a business, its website and market. Requires the business's own website (not a social profile or link-in-bio page); ask the user for it before calling; without one, offer the direct creator, video or search actions instead. Uses normal login, including consumer email. This is optional: a specific creator, video or export request should use its own action directly. Returns a durable operation; read candidate themes, continue with stored IDs, then retrieve the complete report: submission or partial evidence is not completion. Failed or partial-only reports restore eligibility; reconnects for the same brand, market and platform reuse released work. If the free check is already allocated to a different brand, market or platform, status is complimentary_in_use: nothing new was started, so tell the user which brand and market holds it (operationId, locator, market) and never present that report as the requested one.

Continue the same free check through deep collection using 3–5 candidate theme IDs and verified competitor IDs from its latest snapshot. Preserve the requested brand and market. Select clear matches directly; ask the user only when intent is ambiguous. Never invent theme IDs or snapshot versions.

Retrieve progress, candidate themes or the released complete Visibility Check. Read the same operation after reconnecting; polling costs nothing. Return evidence, scope, freshness and limitations honestly. A report awaiting review is not yet released, and a partial report is not a complete outcome.

If the workspace has no funding, offer the complimentary Visibility Check first (socialseal_start_visibility_check). When this action returns reason funding_required with suggestedTool, it did not run: relay its text, collect what missing lists, and once the user agrees call that tool with suggestedArguments plus brand as locator, market as market and the business's own website as targetWebsiteUrl. Start a bounded video search for a query, platform and region. Subscription workspaces pay credits; prepaid workspaces pay the returned USD price only when the search completes with results. Returns jobId when admitted, or a saved operationId with status funding_required when the balance is short: when that result carries suggestedTool, relay its text and offer the free check; otherwise, after a top-up, send the same request again with the same idempotency key; never replace the pending request. Poll socialseal_get_video_search for admitted legacy jobs. Concurrent equivalent searches coalesce. No ongoing tracker is created. Preserve returned coverage and provider limitations. Search rows describe videos surfaced for selected queries at capture time, not search volume or total market demand. Within one platform, market and capture window, result counts, rank and views remain valid relative demand signals for comparing queries; when asked to prioritise, rank with them and label confidence instead of declining. Region is the collection market, not verified viewer geography. Views/likes/saves are video engagement snapshots, not engagement attributable to that query or gains during the export window. State query/platform/market/date scope and coverage; deduplicate videos before aggregating across queries. Define rate denominators and preserve missing values as unknown. Inspect relevant video evidence before recommending a creative treatment.

Read status, available results or a typed failure for an existing video search. Pass its canonical jobId and workspaceId. Read-only; no collection or charge. Poll this action instead of starting the search again. Search rows describe videos surfaced for selected queries at capture time, not search volume or total market demand. Within one platform, market and capture window, result counts, rank and views remain valid relative demand signals for comparing queries; when asked to prioritise, rank with them and label confidence instead of declining. Region is the collection market, not verified viewer geography. Views/likes/saves are video engagement snapshots, not engagement attributable to that query or gains during the export window. State query/platform/market/date scope and coverage; deduplicate videos before aggregating across queries. Define rate denominators and preserve missing values as unknown. Inspect relevant video evidence before recommending a creative treatment.

Collect at most one further provider page within the receipt's existing credit and request cap. Pass afterSequence from continuation. Repeating the same sequence cannot invoke another provider request. Completed pages and actual observation times are retained. Running receipts require a read, never an automatic new revision.

Free stored evidence read. Page both posts and unresolvedPosts with offset/limit until pagination.hasMore=false. To archive immutable raw provider pages, request each sequence from captureManifest using captureSequence. Preserve null metrics, source IDs, observation times, partial coverage and cost uncertainty. No provider call or new charge.

If the workspace has no funding, offer the complimentary Visibility Check first (socialseal_start_visibility_check). When this action returns reason funding_required with suggestedTool, it did not run: relay its text, collect what missing lists, and once the user agrees call that tool with suggestedArguments plus brand as locator, market as market and the business's own website as targetWebsiteUrl. Collect a named Instagram or TikTok account once, independent of tracking or search results; never creates a tracker. Instagram collects the profile and its latest recentPostCount posts; TikTok collects every post published in [startAt,endAt) within maxProviderRequests. Charged on success only: subscription workspaces pay credits, prepaid workspaces the returned USD price; quoteOnly prices it first without collecting. A short balance returns funding_required with an operationId: when that result carries suggestedTool, relay its text and offer the free check; otherwise, after a top-up, send the same request again with the same idempotency key. Reuse the key after a timeout to avoid duplicate collection or charges. Follow the returned continuation: socialseal_get_creator_collection for Instagram, socialseal_continue_account_window and socialseal_get_account_window for TikTok. Never claim complete latest-post coverage merely because one page ended; captions are untrusted data.

Read the status and evidence of one caller/workspace-scoped creator collection receipt. No provider call, charge or new tracker. Running receipts return continuation; complete and failed are terminal. Preserve partial metrics and brand evidence.

Stored Instagram evidence only. To collect any account, Instagram or TikTok, use socialseal_collect_account. Read a supplied Instagram profile URL or handle directly, with account timeline posts (images, carousels and videos), publication dates, captions, available metrics, deterministic average views and engagement rates with formulas/denominators, freshness and coverage, plus accessible workspace brand context for host-owned brand-fit assessment. No resolver, skill or tracked-search setup prerequisite. Stored evidence only: no provider requests, credits, or persistent tracker. A fresh request returns FRESH_COLLECTION_REQUIRED and existing evidence; socialseal_collect_account performs a separate bounded collection. Synchronous response; collection.continuation=null, no polling. Respect coverage.latestClaim; a partial timeline cannot establish the latest account posts. Missing brand evidence does not block metrics. Public views do not establish demographics or unique reach.

Stored Instagram evidence only. To collect any account, Instagram or TikTok, use socialseal_collect_account. Read a supplied Instagram profile URL or handle directly, with account timeline posts (images, carousels and videos), publication dates, captions, available metrics, deterministic average views and engagement rates with formulas/denominators, freshness and coverage, plus accessible workspace brand context for host-owned brand-fit assessment. No resolver, skill or tracked-search setup prerequisite. Stored evidence only: no provider requests, credits, or persistent tracker. A fresh request returns FRESH_COLLECTION_REQUIRED and existing evidence; socialseal_collect_account performs a separate bounded collection. Synchronous response; collection.continuation=null, no polling. Respect coverage.latestClaim; a partial timeline cannot establish the latest account posts. Missing brand evidence does not block metrics. Public views do not establish demographics or unique reach.

Find existing brand facts, constraints and decisions in authorized workspace notes. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read an existing note by its returned ID. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read core context, building its stored cache from existing notes if absent. No provider collection or model cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Save a requested fact, preference, decision or constraint. No provider collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Replace the current workspace note with the supplied content as a new version, mark the previous version superseded and rebuild core context. Prior versions remain stored. Returns the new note and superseded note identifier. No provider collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Change whether the specified note is included in core context. No provider collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Rebuild the core context from existing notes; does not collect social content. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read the published content Strategy for a Brand (default: the Workspace's primary Brand): offer, wanted and unwanted audiences, frictions, messaging and production rules, angle ledger and decision tests, plus any pending draft. Call before content strategy, ideas, briefs, calendars, captions or interpreting demand/performance data, and apply it as described in its guidance. state=missing (no Strategy yet) and state=no_brand (no primary Brand yet) are normal results, not failures: proceed with the user's request on stated assumptions and follow the returned guidance. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Save a draft Strategy revision for user review. Pass the complete Strategy and expectedBaseRevisionId (the published revision you read, or null when none exists). Drafts do not affect agents or generation until published. Use it when the user asks to create or change the Strategy, or after they confirm a correction that should persist. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Publish a draft Strategy revision so agents and brief generation apply it. Call only after the user has explicitly approved this exact draft in the conversation. Fails with BRAND_STRATEGY_STALE_BASE when the published Strategy changed after the draft was proposed. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Retrieve existing brand context before asking the user for brand identifiers. Returns canonical brands and workspace roles. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Create the explicitly requested brand identity and workspace role; no account monitoring or collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read named sets of existing brands. These are not tracking groups. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Create a named brand comparison set; no tracking or collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Update a comparison set's name or description; no collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Delete the requested comparison set; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Add an existing brand to the requested set; no tracking or collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Remove membership in the requested set; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read known attribution aliases for an accessible brand. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Attach a handle, hashtag or keyword attribution alias. A handle alias marks that platform account as brand-owned and requires its platform; it does not monitor the account; no collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Change the active state of one attribution alias; no collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Ask the specified queries of one engine, one run per query: google_ai (Google's AI Overview, the default), chatgpt, gemini or claude. Each answer is kept with the pages it cites and the brands it names. A Claude query costs twice a Google AI, ChatGPT or Gemini query; quote first with quoteOnly. Returns runIds (one numeric runId per query; runId is the first); poll socialseal_get_tool_status with kind=google_ai_run for each and read results after completion. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read Google AI, ChatGPT, Gemini and Claude run status and progress, scoped to the requested run or tracking item. Each run names its engine. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read stored Google AI, ChatGPT, Gemini and Claude answers and citations, preserving the actual query and historical capture scope; each answer's run names its engine. Filter with engine to compare one engine's answers. Paginate offset/limit. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Create the requested single-platform tracking group; does not establish populated or fresh evidence. A group with the same name (trimmed, case-insensitive), platform and locale_key is returned instead of a duplicate, with reused_existing=true. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Merge the requested source groups into the target group: the target keeps the union of their items (a shared item is still collected once) and the source groups are deleted. All groups must use the target's platform; a source group still referenced elsewhere blocks the whole merge without changes. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Delete the explicitly listed groups; their tracking items stay in the workspace. Reports deleted, missing and blocked group IDs separately. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Attach existing item IDs, or add search values with Topic assignments. Search adds use canonical Topic resolution; omitted Topics never create an unclassified active search. Returns item-level or typed activation outcomes. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Remove only the requested group membership. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Update the requested group name, description or refresh schedule; scheduling can affect future collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Delete the explicitly requested group; does not silently replace it or recollect evidence. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Queue collection for the existing group. May incur collection cost. Returns the existing refresh polling contract; no completion or freshness claim before status/evidence confirms it. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read the existing group's collection queue state; does not start another refresh. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Show existing authorized tracking groups across their configured platforms. Reads stored configuration without collection, refresh or setup; no new collection cost. Use the returned group ID and workspace for subsequent reads. Groups overlap, so never add up items_count for a tracked total; use socialseal_get_tracking_inventory.

Read an existing tracking group's configuration by ID. Reads stored data; does not refresh evidence or create tracking. No new collection cost.

Compare expected memberships with stored items in the authorized tracking group, including missing/extra items, platform mismatches and optional stored refresh-job status. Omit expected_items to inspect current membership. Reads existing configuration/jobs only; no collection, refresh, mutation or new cost. Completeness does not establish fresh evidence. Reuse the workspace from the group lookup.

Read the monitored items in an existing tracking group, paginated by page and limit. Stored membership is not fresh search evidence. No collection or new collection cost; no setup prerequisite.

Read the workspace's canonical tracked-item total and breakdown: one item per search in one market on one platform, hashtag or account, counted once however many groups include it, with paused items separate. Use this total whenever you state how many searches, hashtags or accounts are tracked; do not add up group item counts or count a page of list results. Pass includeItems for a page of items, including paused items, with platform, market, frequency, last and next collection, latest refresh outcome and Monitor report inclusion. The summary always covers the whole workspace. To read every item, repeat with offset=pagination.nextOffset while pagination.hasMore is true. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read existing tracked items and their actual platform/market scope. Results are paged; for how many items are tracked, use socialseal_get_tracking_inventory. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read one existing tracked search, hashtag or account. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Find a target by exact type, value, platform and region, including inactive matches. Returns found=false when it is not tracked in this workspace. Does not create tracking. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read active Topics with stable IDs, names and descriptions before assigning searches. No collection cost. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Create only the explicitly requested ongoing account commitment. May schedule provider refresh and incur collection cost under existing controls. Use creator reads for a one-off evaluation. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Create the explicitly requested search/hashtag commitment. Active searches require Topic resolution: pass topicDefinitions and topicRefs, or omit them for automatic organisation. Unambiguous single-search automatic results preview and execute together; ambiguous results return TOPIC_INPUT_REQUIRED without activating. May schedule collection and incur cost under existing controls. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Return the exact Topics to create or reuse, per-search assignments, conflicts, credits and planFingerprint without activating. Use execute_search_activation with that fingerprint to commit. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Commit Topics, assignments and search activations atomically for a previewed plan. Requires planFingerprint and idempotencyKey. Duplicate retries return the original created or reused result. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Request fresh collection for an already monitored item. May incur collection cost; retain the returned job/status information and do not duplicate starts. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Resume collection for a paused item. The result's collection field says what will collect it: monitor, prepaid or subscription (collected from the next scheduled run; prepaid runs are paid from Workspace funds and wait when funds are short), or manual (only on request). Items whose platform cannot collect in their region (TikTok in CN or HK) stay paused and return UNSUPPORTED_PLATFORM_REGION; report that reason rather than retrying. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Update the requested item configuration; scheduling changes can affect future collection cost. is_active: true resumes a paused item (same rules as socialseal_resume_tracking_item); pausing is not supported. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Delete the requested tracking item. This changes ongoing monitoring; it is not needed for account metrics. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read stored analysis for returned video identifiers. URLs are not accepted for polling; missing or pending evidence does not trigger provider work. Separate observed hooks/shots/transcripts from model interpretations. Photo/carousel analysis uses one-based slideIndex references; it provides no video timing, motion or audio evidence. likelyReasonItRanks and similar explanations are hypotheses, not measured ranking causes or independent proof of effectiveness. Cite the source post and inspect whether its audience, topic and production context transfer to the requested brief. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

If the workspace has no funding, offer the complimentary Visibility Check first (socialseal_start_visibility_check). When this action returns reason funding_required with suggestedTool, it did not run: relay its text, collect what missing lists, and once the user agrees call that tool with suggestedArguments plus brand as locator, market as market and the business's own website as targetWebsiteUrl. Retrieve assets for stored post identifiers. TikTok photo/carousel posts return ordered signed slide images; video posts return video/frame assets. Missing frame extraction or requested analysis may incur processing cost under existing controls. Returns per-item state and source links; poll analysis by returned ID, not URL. Separate observed hooks/shots/transcripts from model interpretations. Photo/carousel analysis uses one-based slideIndex references; it provides no video timing, motion or audio evidence. likelyReasonItRanks and similar explanations are hypotheses, not measured ranking causes or independent proof of effectiveness. Cite the source post and inspect whether its audience, topic and production context transfer to the requested brief. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

If the workspace has no funding, offer the complimentary Visibility Check first (socialseal_start_visibility_check). When this action returns reason funding_required with suggestedTool, it did not run: relay its text, collect what missing lists, and once the user agrees call that tool with suggestedArguments plus brand as locator, market as market and the business's own website as targetWebsiteUrl. Queue analysis of a supplied public TikTok, Instagram or YouTube video URL, or a TikTok photo/carousel post, without creating persistent tracking. Requires an authorized workspace; retrieve accessible workspace context before asking for an ID. May incur collection/analysis cost under existing backend admission controls; preserve required approval. Photo analysis is priced per image and returns ordered slide references rather than video timestamps or audio evidence. Returns per-item status and identifiers; poll socialseal_get_tool_status with kind=video_analysis and the returned videoUid or platformVideoId, never the URL. Private/unavailable posts and provider errors are limitations, not a reason to set up tracking. Do not assume fresh collection or successful completion.

Read existing source clips, rights metadata and available clip analysis; expiring media URLs are not permanent assets. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Create the requested source clip upload and return signed upload fields. Requires a separate byte upload and finalize action; no assembled-video output. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Finalize the requested upload only after bytes are uploaded and rights are attested. May trigger existing clip-analysis processing; preserve provider/admission failures. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Start bounded keyword expansion for the specified subject/market. May incur collection/analysis cost under existing backend controls. Returns runId; poll socialseal_get_tool_status with kind=journey_run. Creates an execution, not an ongoing tracker. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read status and available results of the returned journey run; do not restart it to poll. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read the requested version and source evidence of a saved blueprint; missing evidence remains explicit. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Save supplied evidence-backed blueprint content without provider collection. Do not fabricate evidence citations. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Generate a blueprint for the supplied scope or manual videos. May consume credits; use previewOnly to inspect selection and retain existing admission/readiness controls. Read the returned opportunity/version with socialseal_get_blueprint; only a usable generated blueprint should feed socialseal_generate_brief. Draft means pending evidence and missing_data requires inspecting the stated gap. This is reference synthesis, not measured proof of creative effectiveness or a finished video. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read existing shot panels and source references with expiring signed URLs; no new collection. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Regenerate reference panels for the selected blueprint. May incur processing cost under existing controls; this does not assemble a video. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read an existing brief version and evidence by opportunityId (or the deprecated opportunityKey); the response returns opportunityId. Brand/strategy setup is not required to retrieve it. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Save the requested creator-brief edits as a new version without modifying or deleting prior versions. Unspecified fields, evidence lineage and pinned context are preserved; supplied arrays replace that field in the new version (null arrays clear it). Supply briefId or opportunityKey. A stale base version is rejected. Reuse editId on retry; an unchanged retry can return NO_CHANGES before replaying the original result, so read the saved brief to verify its version. Without editId, retries may create additional versions. A successful write returns the saved brief, version and deduplication status. No collection cost. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Save supplied brief content tied to a real blueprint version; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Create a video, creator, UGC or editor brief with hooks and shot direction. Use an existing blueprintId/version or a topic, competitor or tracking_group scope; retrievalPrompt refines scoped retrieval and cannot accompany blueprintId/version. Carry approved audience/product facts and constraints in brandContext; do not assume project documents are loaded. Read existing work with socialseal_get_brief, edit with socialseal_update_brief, and export with socialseal_export_brief before formatting as Word/PDF. May consume credits under existing controls. Generated status does not establish factual accuracy, executable timing or creative quality; inspect the returned evidence and missing-data states. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Export the requested brief version for handoff. No new provider collection; preserve returned artifact and evidence links. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Check evidence readiness for creative brief generation or editing. This check does not gate unrelated reads, account metrics or URL analysis. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Start the requested topic's brief execution, preserving backend readiness, budget and idempotency. May consume credits. Poll socialseal_get_brief_run with the returned runId; no social publishing or persistent account tracker. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read progress, result and failure evidence for an existing brief run. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Retry the existing run under backend context-fingerprint and idempotency checks. May consume credits. Poll socialseal_get_brief_run; do not silently force a full rerun. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read outcomes only for posts actually linked to the specified brief: per post, the Monitor's Share of Voice and Discoverability for the Brief's topic before and after publication (monitor, with method and missingReasons; a post absent from tracked results is not_observed and unknown views stay null), plus legacy weekly search rank as a secondary line. Observational, not causal; this sample is not total account reach. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Attach an externally published video that this saved Brief produced. Requires the exact Brief ID, not a reference video. Reuses canonical video identity; duplicate conflicts preserve the original binding and baseline. URL resolution may incur collection cost under existing backend billing controls. Omitted trackingItemIds explicitly selects all workspace tracking; provide target IDs and region for scoped evidence. This does not publish content or prove causal uplift. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

List existing journey identities and optionally their latest runs. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read a selected existing journey and optionally its latest execution. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read diagnostics for the existing journey scope without starting another run. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Read existing keyword records and their evidence/strategy associations. Each keyword carries demandSignals: evidenceLevel (strong/indicative/untested), the demand signals present, and googleSearchVolume.state (not_collected means never fetched, not zero). Listing keywords does not start searches. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Rank candidate keywords, hooks or titles by demand evidence before choosing a wording. Checks live Google and YouTube autocomplete (free public endpoints; no SocialSeal charge) and reads Google search volume already stored for the workspace. Returns results ranked by evidenceLevel (strong/indicative/untested) with autocomplete state (exact/variant/extended/absent/failed), position and observed alternative wordings. Prefer an evidenced alternative over an untested wording. When ranking, also weigh Google volume and trend, surfaced-result response and engagement, and Explore relative-demand components where available; missing Google volume is unknown, not absent demand, and wordings still without a signal rank last. Does not fetch new Google Ads volume or TikTok autocomplete; use socialseal_start_search_journey or Explore for those. Saves nothing.

Save the specified keyword and associations; does not create monitoring or collect provider data. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Save the specified keyword batch; does not create monitoring or collect provider data. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Update the requested keyword record and associations; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Apply the requested patch to the selected keyword IDs; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Delete only the requested keyword record. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Copy existing tracked keyword associations into the keyword registry; does not run provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Read saved workspace personas. Personas are supplied strategy records, not inferred private audience demographics. Reads stored evidence/configuration; no provider collection or collection charge. Preserve returned timestamps, coverage, missing evidence and errors. Synchronous; no polling unless the result explicitly supplies continuation.

Save the user's evidence-backed strategy persona; never fabricate private audience demographics. No provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Apply requested changes to a saved strategy persona; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Mark the requested strategy persona inactive; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

Reactivate the requested existing strategy persona; no provider collection. Changes only the requested workspace state. Backend authorization, validation, pricing, balance and idempotency controls remain authoritative; tool selection does not authorize additional work.

List workspaces visible to the authenticated SocialSeal user. List order and owner role do not identify the default; use defaultWorkspaceId/isDefaultWorkspace, or call socialseal_get_current_workspace before a write when the user did not name a workspace.

Get the current workspace before a workspace-scoped action when the user has not named one. Hosted MCP returns the sole workspace the user can reach, or none when there are several: then ask the user which one. Local MCP returns the workspace selected in environment/configuration.

Takes no parameters.

Read-only polling for a known SocialSeal async run, export, or group refresh. UUID auto ids are journey runs; numeric auto ids are AI answer runs (Google AI, ChatGPT, Gemini or Claude). Both are workspace-scoped, as are google_ai_run, journey_run, explore_run, and video_analysis. Pass the workspaceId from the lookup that returned the run. MCP-host errors such as "No approval received" occur before SocialSeal executes this tool; never describe them as SocialSeal billing, workspace approval, or job-status errors.

Read the server-owned acquisition plan for an authorized workspace, anchor, market and language. Returns planFingerprint, estimated credits, capacity, source availability, blocking reasons and confirmationRequired. Does not create a lineage, queue a run or collect provider evidence. Pass the returned planFingerprint as previewFingerprint to socialseal_explore_start with the same scope. Ordinary plans need no additional user confirmation; request scope confirmation only when confirmationRequired is true.

Start audience-demand research for the supplied anchor, market and language. Creates a research lineage and queues a collection run, which may consume credits and provider costs. First call socialseal_explore_preview_start; use its planFingerprint as previewFingerprint with the same scope, explicit workspace and retry key. Returns lineage/run identifiers and execution state; inspect the run for results. If the server requests scope confirmation, supply confirmedScopeFingerprint only after the user confirms that scope, reusing the same idempotencyKey.

Read the canonical workspace Explore overview, including lineages, latest snapshots, recommendations, capabilities, and blocking reasons. Each snapshot carries signalRanking: its terms ordered by observed demand signals with an evidenceLevel (strong/indicative/untested), usable for prioritising even when the calibrated index is null. The default summary view omits evidence bodies; read a snapshot with socialseal_explore_get for full components and evidence.

Read one canonical Explore lineage, snapshot, run, selection, or compatible snapshot comparison. The response is returned without collapsing evidence, unavailable sources, or partial states. A snapshot's signalRanking orders its terms by observed demand components (Google volume, autocomplete salience, search-result response and engagement); a null calibrated index or insufficient_evidence state means the composite was gated, not that demand is unknown, so prioritise with signalRanking and its evidenceLevel.

Queue a new demand-research run for an existing Explore lineage. May collect new provider evidence and consume credits. Requires the workspace, lineage and retry key. Returns run identifiers and execution state. Reuse the same idempotencyKey; supply confirmedScopeFingerprint only after the preview required confirmation (an unusually large scope, or a scheduled search moving to another cadence).

Persist an exact, user-confirmed Explore selection for a stated destination and intended use. No additional terms are inferred or added.

Apply one canonical interpretation correction with stable retry identity and the exact labels and memberships supplied by the caller.

Preview tracking an immutable Explore selection's searches under one Topic on the chosen platforms and cadence, without tracking anything. Returns the server plan, its planFingerprint, searches already tracked on another cadence (cadenceChanges) and whether confirmation is required.

Track the selected Explore searches in Monitor under one Topic on the chosen platforms and cadence, using the planFingerprint from the activation preview. Reuses matching tracked searches and moves them to the chosen cadence; a paused one must be reviewed first. Setup is free; the first scheduled refresh starts on the next scheduler pass and each refresh is charged when it collects. Returns status, Topic and tracking item ids; collected results are available later. Requires an explicit workspace and idempotencyKey; supply confirmedScopeFingerprint only after the preview required confirmation (an unusually large scope, or a scheduled search moving to another cadence).

Prepare a server-issued Explore continuation for the selected destination. A handoff is not action acceptance and carries no client-inferred payload.

Create a shareable JSON report from an existing Explore selection, including its saved search terms, interpretation, evidence and limitations. Saves an immutable report and creates an external share link valid for 30 days. Requires an explicit workspace, selection and idempotencyKey. Retries reuse the report but issue a fresh share link.

Export legacy time-window tracking data for the authenticated SocialSeal user. Always reuse the workspaceId from the lookup that returned the selected group or tracking item; numeric IDs are not sufficient workspace context. Fulfilled as an export artifact: returns download_url plus a read_token for in-session paginated reads with socialseal_read_export_chunk when the URL cannot be fetched. For ranked discoverability, use socialseal_export_report with reportType search_results_enriched.

Export a SocialSeal report. Use reportType search_results_enriched, format csv and payload.groupIds for ranked search results; reuse the workspace from the group lookup. Poll the returned export ID while pending, then fetch download_url or page socialseal_read_export_chunk with read_token until has_more is false. Report partial coverage if retrieval stops early. Search rows describe videos surfaced for selected queries at capture time, not search volume or total market demand. Within one platform, market and capture window, result counts, rank and views remain valid relative demand signals for comparing queries; when asked to prioritise, rank with them and label confidence instead of declining. Region is the collection market, not verified viewer geography. Views/likes/saves are video engagement snapshots, not engagement attributable to that query or gains during the export window. State query/platform/market/date scope and coverage; deduplicate videos before aggregating across queries. Define rate denominators and preserve missing values as unknown. Inspect relevant video evidence before recommending a creative treatment.

Read rows of a previously generated SocialSeal export artifact in-session, paginated as JSON. Use this when the export tool returns a download_url you cannot fetch (e.g. sandbox egress limits or web_fetch cannot render a CSV attachment). Pass the short download token (the `t` query param from download_url) plus offset and limit. Each call returns { export_history_id, headers, rows, offset, limit, next_offset, total, has_more }. Rows are packed by serialized char budget (not a fixed row count), so `limit` reflects the actual number of rows returned and may be smaller than requested. To page: if has_more is true, call again with offset = next_offset. Repeat until has_more is false. Prefer limit=200 as the max-rows cap. The data stays inside the MCP tool channel, so no URL fetch is required.

Discover the canonical SocialSeal concept registry: metric concepts (discoverability, visibility, share of voice, reach, ranking, sentiment) and the domain glossary, with definitions, distinctions, required dimensions, evidence requirements, and input-only aliases.

Look up released capability metadata. Optional reference; this does not gate execution or authorize work.

Discover optional task-specific SocialSeal guidance after identifying the requested work. For a content calendar or posting schedule, select socialseal-social-plan-builder; for video/editor/UGC briefs, select socialseal-creator-briefing. Loading guidance does not execute generation tools. Skills are not prerequisites for basic operations. Returns each skill's name and frontmatter description (so you can route without reading every SKILL.md), file paths, shared references, the install command, and the archive download info.

Takes no parameters.

Read the full instructions for one SocialSeal skill. Provide skill (and optional file, defaulting to SKILL.md) to read within skills/<skill>/, or provide path to read a repo-relative file under skills/, references/, docs/, templates/, or examples/.

Return download URLs for the full SocialSeal agent-skills repository (zip/tarball) plus the install command, so the agent or user can download every skill at once.

Takes no parameters.