StackFast FractWin Expert Brain

Return session artifact metadata and persistent URLs. V1 intentionally omits raw artifact contents.

Discover wallet-resolved credential service names and accepted aliases without exposing secret values. Use this when an agent is unsure whether a key exists, sees a key-not-found error, or needs the canonical getAgentKey(service) name. Returns service slugs, env/key aliases, categories, and resolver guidance only; never returns raw credentials.

Dealership OS checks a tenant-scoped lead against consent, STOP, quiet-hours, pacing, verified-sender, identity, assignment, and manager-exception rules. It returns an eligibility or hold decision, a human action card, and an opaque receipt. The entitled customer modes are evaluate and synthetic_acceptance. It never sends outreach, clears STOP, writes CRM or inventory, reserves inventory, calculates a deal, approves desk/OTD terms, or replaces human authority; governance registration and STOP recording are operator-only modes.

Dealership OS returns a public-safe, tenant-scoped coaching brief for opportunity leak diagnosis, ownership and next actions, daily manager reconciliation, customer connection, customer-first presentation structure, or an opportunity disposition recommendation. It uses the governed T0 public judgment rail and returns findings, an ordered plan, next actions, citations, a human action card, and an opaque receipt. It never sends outreach, writes CRM or inventory, changes opportunity authority, approves desk/OTD terms, performs an assessment, grants a credential, or exposes protected method bodies.

Read AI Stack Audit project state, deliverable refs, credit ledger, and Desk/CRM links.

Generate a governed, tenant-scoped artifact draft with no-fabrication placeholders, analyzer provenance, MASK, brand status, and approval requirements.

Refine an existing artifact draft while preserving tenant boundary, provenance, MASK status, and approval posture.

Approve/save an artifact and mark tenant-scoped learning-loop eligibility. Does not publish externally.

Create an audited export receipt for an approved artifact or internal draft. External publish remains separate and human-approved.

Compose a relationship-safe correspondence decision packet with do_not_say, safe_phrasing, channel constraints, manual approval, and a correspondence_response artifact audit trail. It never sends the message.

Read provenance, analyzer findings, gate status, approval state, usage, and export receipts for one tenant-scoped artifact.

Build a private GrowthOS Revenue Capture Scorecard from supplied public business signals. Human buyers start at /growthos/scorecard; MCP execution requires reader entitlement. Returns scorecard findings only; it never sends outreach.

Start or append a GrowthOS Business Brain Interview using the existing interview/session model. Outputs remain draft_review_required before automation.

Return a GrowthOS revenue-capture packet: scorecard, review-gated contact drafts, CogentCast site-review receipt composition, and summary for supplied local-business evidence. GrowthOS consumes CogentCast receipts instead of recreating website-review logic. No autonomous outbound.

Render the canonical GrowthOS scorecard into a sendable owner-facing report with web, email-review-ready, PDF-ready outputs, and optional CogentCast site-review receipt composition. No autonomous send.

Build an Opportunity Scout packet for customer, talent, contract, or capital opportunities using the existing gate-first policy layer and no new persistence table.

Build a draft GrowthOS digital-clone render packet from normalized signals, evidence, runtime tenant voice profile, and PERSPECTIVE capability grounding. No Robert voice defaults and no autonomous outbound.

Validate a Work Sample, content_type=narration_review, or content_type=audiobook_production configuration without creating state or vendor cost. A failed preview returns no creation-authorizing digest or authoritative cost estimate. Audiobook Production v1 reuses this five-tool surface: content_type=audiobook_production registers exact manuscript/script/SSML plus provenance, pronunciation, visual/companion, and deterministic-build receipts at zero cost; production.status returns the current derivative and Chief Editor/Gemini/ear/paid/publication gates. Paid canary, full generation, and publication are distinct receipt-gated stages. Uses the existing CogentCast production service; it does not bypass Studio, mint Chief Editor/Robert approval, substitute model output for human authority, publish publicly, or expose credentials.

Create one governed Work Sample from a tenant-bound preview reference, preserve the private narration-review create contract, or append a zero-cost audiobook preparation record through CHIEF_EDITOR_REVIEW. No create path can mint a human pass or make a vendor call. Audiobook Production v1 reuses this five-tool surface: content_type=audiobook_production registers exact manuscript/script/SSML plus provenance, pronunciation, visual/companion, and deterministic-build receipts at zero cost; production.status returns the current derivative and Chief Editor/Gemini/ear/paid/publication gates. Paid canary, full generation, and publication are distinct receipt-gated stages. Uses the existing CogentCast production service; it does not bypass Studio, mint Chief Editor/Robert approval, substitute model output for human authority, publish publicly, or expose credentials.

Request one leased production stage. The zero-cost Gemini reconciliation stage requires exact dispatch, task, job, source, and Chief Editor receipt binding and cannot mint verification. Cost-bearing canary/full-chunk stages require exact prior Chief Editor, Gemini, and human receipts; provider ambiguity blocks paid retry. This tool cannot record human approval or publish. Audiobook Production v1 reuses this five-tool surface: content_type=audiobook_production registers exact manuscript/script/SSML plus provenance, pronunciation, visual/companion, and deterministic-build receipts at zero cost; production.status returns the current derivative and Chief Editor/Gemini/ear/paid/publication gates. Paid canary, full generation, and publication are distinct receipt-gated stages. Uses the existing CogentCast production service; it does not bypass Studio, mint Chief Editor/Robert approval, substitute model output for human authority, publish publicly, or expose credentials.

Read tenant-scoped Work Sample, narration-review, or audiobook-production state; exact hashes; current-vs-historical derivative projection; and Chief Editor/Gemini/ear/paid/publication gates. Private source bodies and blob locations are not exposed. Audiobook Production v1 reuses this five-tool surface: content_type=audiobook_production registers exact manuscript/script/SSML plus provenance, pronunciation, visual/companion, and deterministic-build receipts at zero cost; production.status returns the current derivative and Chief Editor/Gemini/ear/paid/publication gates. Paid canary, full generation, and publication are distinct receipt-gated stages. Uses the existing CogentCast production service; it does not bypass Studio, mint Chief Editor/Robert approval, substitute model output for human authority, publish publicly, or expose credentials.

Read and validate a server-recorded, human-authenticated approval receipt by id. An LLM cannot create approvals, impersonate Robert, change visibility, or unlock public release. Uses the existing CogentCast production service; it does not bypass Studio, mint Chief Editor/Robert approval, substitute model output for human authority, publish publicly, or expose credentials.

Operate on tenant-scoped OCE Estimator/Appraiser catalog sources/items, including the StackFast storefront catalog, through the generic estimator_* control-plane contract.

Operate on tenant-scoped OCE Estimator/Appraiser catalog sources/items, including the StackFast storefront catalog, through the generic estimator_* control-plane contract.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Prepare an OCE Estimator/Appraiser document or QBO handoff contract. Review-gated; does not write to accounting systems.

Prepare an OCE Estimator/Appraiser document or QBO handoff contract. Review-gated; does not write to accounting systems.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Operate on tenant-scoped OCE Estimator/Appraiser drafts, lines, versions, documents, exports, and policy receipts.

Read the Desk ticket, CRM callback task, call-link receipt, and A2P-gated inbound SMS posture for one AI Receptionist call ID. Returns masked/status metadata only.

List recent AI Receptionist call-loop receipts for human review, including callback and A2P-gated inbound SMS status where present. Returns masked/audit-safe identifiers only.

Return a human-review callback packet for one AI Receptionist call. The packet is review-only and does not send SMS, place calls, or reply to customers autonomously.

Create or update the account owner's private Talent Scout profile setup once so compose, fit, and packet tools can reuse self-entered resume, LinkedIn, proof-point, preference, and voice evidence without re-pasting it for every role. Existing profiles require the current profile_version_updated_at from Talent Scout My Profile Status as if_match_updated_at. Stores private owner evidence for proof mapping and never returns raw resume or LinkedIn text.

Create or update the account owner's private Talent Scout profile setup once so compose, fit, and packet tools can reuse self-entered resume, LinkedIn, proof-point, preference, and voice evidence without re-pasting it for every role. Existing profiles require the current profile_version_updated_at from Talent Scout My Profile Status as if_match_updated_at. Stores private owner evidence for proof mapping and never returns raw resume or LinkedIn text.

Read a compact, redacted readiness receipt for the account owner's private Talent Scout profile setup, including profile_version_updated_at for guarded writes through either public profile-upsert alias. Returns only setup status, compose readiness, and active search-rail readiness; never returns raw resume, LinkedIn, work history, claims, or private owner profile text.

Read a compact, redacted readiness receipt for the account owner's private Talent Scout profile setup, including profile_version_updated_at for guarded writes through either public profile-upsert alias. Returns only setup status, compose readiness, and active search-rail readiness; never returns raw resume, LinkedIn, work history, claims, or private owner profile text.

Correct an owner-reported application-history state with optimistic concurrency and an append-only reconciliation receipt. Use this when an intake or imported record was incorrectly shown as applied. Preserves the history record for audit, removes not-applied records from the active application queue, and never applies, sends, or contacts anyone.

Persist the active Talent Scout search strategy for a tenant: target lanes, compensation floor, geography, positive/negative filters, resume-variant priority, travel posture, and voice anchor. Future search, fit, review, compose, packet, and note calls load these rails by default until superseded.

Create, update, activate, pause, or archive a saved Talent Scout direct-company discovery campaign. Campaigns define company lanes, role lanes, size, geography, comp floor, travel tolerance, negative filters, resume variant, voice anchor, and human-review requirements. Draft/review only; never sends or applies.

Preview the account owner's saved business-list campaign. Returns public company websites, employee-count confidence, source receipts, and next actions for manual review. Read-only by default: it does not persist companies unless persist_results is explicitly true and never scrapes LinkedIn.

Preview the account owner's saved local business-list campaign. Returns public company websites only, with size-band confidence and source receipts for manual review. Read-only by default; set persist_results true only for an intentional store update.

Preview a qualified company target for direct careers URLs and role-lane matches. Returns verified status receipts such as company_careers_open_role, needs_human_url_verification, needs_jd_hydration, or careers_page_no_relevant_role. Read-only by default: it returns exact role records for human review, does not persist queue rows unless persist_results is explicitly true, and never invents jobs.

Generate public contact-research labels and safe LinkedIn search queries so the account owner can decide who to review manually. No LinkedIn scraping, no auto-connect, no auto-message; human review required.

Prepare Hunter.io-style business email enrichment receipts for a qualified contact. Uses wallet SSoT status, verifies before recommending, never sends, and requires human approval.

Preview a manual note for Robert to copy or edit himself. This read-only tool does not send, store, contact anyone, scrape LinkedIn, enrich private people data, or create a draft record.

Record a human decision and receipt on a Talent Scout manual-note draft. This is closed-world tracking only: it never sends email, LinkedIn messages, phone calls, or applications automatically. Use sent_text only to store text the human already handled manually.

Record one human-reported email, LinkedIn, SMS, call, or visit in the existing StackFast Relationship Core. Returns an immutable, idempotent receipt and makes follow-up state visible in Talent Scout. It never sends, scrapes, applies, submits, or changes candidacy status.

Append an immutable correction for the channel and direction of an existing Talent Scout Relationship Core receipt. The original receipt is never changed; this tool never sends, applies, scrapes, or changes candidacy status.

Record that the human manually submitted an application. A known Talent Scout queue item is preferred, but a human-attested company, title, requisition/application id, and application URL can create the same durable history-backed queue record when no live posting exists. This updates lifecycle status, schedules follow-up tasks, and records a submission receipt only; it never applies, sends, clicks, or submits anything automatically.

Enumerate every required and optional ATS form field before any fill attempt. Returns a stable schema hash and complete-field receipt; it never types, uploads, applies, clicks, or submits.

Build or verify a keyboard-event ATS fill plan from a complete form schema. It maps safe stored profile values, reserves eligibility/EEO/files/signature/submit for the human, and requires per-field readback before reporting staged.

Record the account owner's manual decision on an existing Talent Scout queue item: keep, top priority, hold/revisit, pass, or not-a-fit. This is closed-world queue bookkeeping only; it never sends outreach, applies, clicks, submits, screens people, ranks people for hiring, or contacts anyone.

Import a guarded BYOC-local Talent Scout queue delta into the cloud queue. This is closed-world queue bookkeeping only: it never applies, sends, clicks, or submits anything automatically.

Return a compact Talent Scout readiness receipt: pipeline version, surface mode, callable tool groups, and manual-review safety flags. Read-only.

Run fresh governed public-opportunity discovery for the account owner to find roles they may review and apply to manually. If live discovery is unavailable or empty, returns fresh_discovery_count: 0 and does not substitute the existing review queue. Draft/review only; never applies or sends messages.

Compare a role text with the account owner's own saved preferences using a deterministic checklist so Robert can decide whether to review or apply manually. Returns transparent self-review data; no outbound action.

Compare a role text with the account owner's own saved preferences using a deterministic checklist so Robert can decide whether to review or apply manually. Returns transparent self-review data; no outbound action.

Explain in prose how an existing saved job overlaps with and differs from the account owner's stated workplace preferences. Returns preference-language observations only, with no numbers, rankings, recommendations, decisions, or external actions.

Create a draft-only application or opportunity packet from an existing Talent Scout queue item. Human approval is required before external use.

Compose a draft-only first-party job-seeker application artifact for the account owner from a real Talent Scout queue item using JD evidence. Supports cover letter, form paste, application Q&A, telephone script, and authority analysis only; never sends, applies, screens people, ranks people, or makes hiring decisions. This does not screen applicants, rank candidates, or make hiring decisions.

Draft a human-reviewed manual note from safe pains and voice evidence. The tool never sends messages and fails closed on auto-send/apply requests.

Read the account owner's own first-party Talent Scout job-search work queue for real synced board roles, draft opportunities, packets, follow-up items, and the daily reconciliation report. This is not an employer review queue, applicant screening queue, candidate ranking system, or hiring decision tool. Mailbox reconciliation runs server-side via the StackFast service-account reader. Clients MUST NOT invoke their own Gmail/email connector; if mailbox_reconciliation_status is not ok, surface the red receipt and stop instead of substituting a client-side mailbox read. Surfaces open roles, direct-URL verification needs, waiting/follow-up buckets, preference filters, and today's slate. Robert's connector resolves to tenant robert and demo rows are quarantined away. Review only; no sends or applications.

Return the closed-loop Talent Scout daily operating report from the shared tenant queue: applied pipeline, mailbox reconciliation status, follow-up due, open roles, availability checks, preference filters, stale/closed rows, new discoveries, bounded needs_role_hydration recovery, and a 3-5 role slate. Mailbox reconciliation runs server-side via the StackFast service-account reader. Clients MUST NOT invoke their own Gmail/email connector; if mailbox_reconciliation_status is not ok, surface the red receipt and stop instead of substituting a client-side mailbox read. Read-only control-tower view; no sends or applications.

Read Talent Scout mail-signal status from the same server-side Gmail reconciliation path used by review_queue. This is the reliable StackFast replacement for local BYOC scout.mail_signals when the cleverq.net/local tunnel is down: it never calls a client Gmail connector, never returns raw mail bodies, and fails soft with mailbox_reconciliation_status instead of a bare 502.

Build the detailed workup for one Talent Scout queue item: availability status, fit risks, company pain map, Robert proof-point map, resume angle, application strategy, and a full draft-only application packet with resume edits, cover-letter outline, form-paste answer, compensation guidance, LinkedIn/follow-up drafts, founder-objection answers, first-30-days plan, questions to ask, claims/risk validation, manual submission receipt template, interview prep, and learning-loop signals. Draft/review only; no sends or applications.

Generate a tenant-scoped, draft-only standalone resume from a hydrated role and stored applicant proof bank, export it to PDF through Artifact Studio, and return a render-readback receipt. It never sends, applies, or approves external use.

Run the voice intelligence orchestrator over any input: an uploaded or recorded audio/video file (base64 in file), a public audio or video URL (audio_or_video_url), a PDF (pdf), an image (image), or raw text/notes (text and/or notes). Provide exactly ONE input carrier — audio_or_video_url, file, pdf, image, or text/notes. The input is normalized to a transcript via the existing transcription rails, then three composed outputs are returned: a condensed summary (with action items, decisions, and a next step), a full-verbatim polished actual script (filler removed, wording preserved, never paraphrased; call sources return a masked summary instead), and an authority report (content flywheel plan, plus an owner-facing business report when business context is provided). All outward outputs are mask-gated.

Search StackFast Brain knowledge through the AI6 MCP reader plane; use fetch with a returned result id for full text.

Fetch a StackFast Brain result or an authorized immutable session artifact by id through the AI6 MCP reader plane. Use the explicit safe shape with id, page, and page_size, for example fetch({ id: "brain:12345", page: 1, page_size: 12000 }) or fetch({ id: "artifact:474", page: 1, page_size: 12000 }); continue with next_page until complete is true.

Read-only StackFast connector health/status check. Does not expose secrets.