Meta Council
Multi-expert decision intelligence with transparent synthesis and auditable workflows.
Community: Submitted by a user or imported; check the owner before granting accessDegradedAPI key requiredGlobalFreeRead-only
What it can do
What data it sees
Do you need an account
An API key from the service settings is required
Multi-expert decision intelligence with transparent synthesis and auditable workflows.
Server tool list (161)
Raw names from tools/list. Only developers need these.
| run_council | Submit a question or decision to Meta Council. A panel of specialized AI agents will independently analyze it, then a synthesis step combines their opinions into a unified recommendation with full transparency. Starts asynchronously by default; use get_session with the returned session id. |
| list_panels | List all available expert panels with their descriptions and agent counts. |
| list_agents | List all available expert agents with their roles, domains, and specialties. |
| get_session | Get the full results of a previous Meta Council session, including all agent opinions and the synthesis. |
| get_visualization | Fetch a chart artifact generated by a council session or LOCUS determination. Returns the machine-readable spec (the data behind the chart) plus the stable SVG URL, or the raw SVG itself with include_svg=true. Artifact ids appear in session results as 'visualizations' / 'visualization' reference blocks. Requires authentication and enforces the artifact owner's tenant boundary. |
| list_workflows | List available multi-step workflow pipelines (composable bundles that chain several steps, each able to run on its own model/provider). Returns each workflow's slug, description, per-step model, and definition_sha — a fingerprint of that definition as loaded right now, comparable against the definition_sha reported by a run you start later. |
| plan_workflow | Preview what a workflow WOULD do, without running it. Costs nothing and runs no models: returns the step execution order, each step's role/model/max_tokens and whether it pauses for a human checkpoint, the workflow's declared parameters (validated if you supply values), which required integrations your account already has credentials on file for, an upper-bound cost estimate, and definition_sha — a fingerprint of the definition this plan was built from, which tells you whether the definition changed between planning and running but never hands back the definition itself. Starts no session and records no usage. Use before run_workflow to check a pipeline fits before spending on it. Requires authentication. |
| test_workflow_step | Run exactly ONE step of a workflow and return what it produced, so you can iterate on a single step's wording without running the steps before it. THIS SPENDS CREDITS EXACTLY LIKE A REAL STEP: the step runs on a real model through the same engine a full run uses and is billed identically -- it is not a simulation, a dry run, or a free preview. If you want to check a workflow's shape, parameters, models and cost estimate for free, use plan_workflow instead; that one runs nothing. Starts no session, so there is nothing to poll and nothing to advance. Requires authentication. |
| run_workflow | Start a multi-step workflow pipeline and return its session id immediately by default. Poll get_workflow_session for each step's model/provider and output. synthesis. Steps run on YOUR configured provider keys, so a pipeline can chain models across providers. If a step is a human checkpoint, returns the session id to advance. Requires authentication. |
| get_workflow_session | Poll a previously started workflow run for its current step-by-step status and output (useful when run_workflow timed out or is awaiting a checkpoint). Requires authentication. |
| recommend_panel | Recommend the best expert panel for a query (semantic match with keyword fallback). Returns the top panel + confidence and the runner-up options — feed the result into run_council's panel argument. Requires authentication because the query may be sent to the configured embedding provider. |
| estimate_cost | Estimate what a panel or saved workflow would cost before running it: returns a per-agent/per-step price breakdown, the total for the requested number of runs, your remaining quota and platform credit, an honest list of any models the platform could not price, and a go/no-go decision. Reserves nothing, charges nothing, starts nothing. |
| get_agent_detail | Full detail on one expert agent by slug (from list_agents): role, description, default model, domain, tags, and tools. |
| advance_workflow | Advance a workflow past a human checkpoint — approve or reject the paused step so the pipeline continues. Use the session id from run_workflow / get_workflow_session. Requires authentication. |
| retry_workflow | Continue a stalled or finished workflow run from a chosen step. This starts a NEW run that reuses the outputs the original run already recorded and only executes — and only pays for — the steps from 'from_step' onward. The original run is left untouched and the new one records which run it continues. Omit 'from_step' to resume at the first step that has no recorded output. A step whose output was never recorded is re-run, never skipped. Requires authentication. |
| get_settings | Get the authenticated user's Meta Council settings — preferred model, plan tier, which credentials are saved (not live-verified), and the registry-backed tools that require, optionally use, or do not use account credentials. Names and booleans only, never secret values. Requires authentication. |
| list_integrations | List which outside services the authenticated user has connected — ticket boards, stored connector credentials (name, kind, whether configured, and when last verified) and the connectors available to call them. Returns configured / not configured facts, names and dates ONLY, and never returns a secret value. Read this before planning a workflow that reaches an external service, so a step is not written against a credential the user has not registered. Requires authentication. |
| get_api_key_info | Report the scopes, plan tier, expiry and remaining quota of the API key making this call, so an agent can check what it is allowed to do before attempting it rather than by being refused. Requires API-key authentication. |
| rotate_api_key | Replace the API key making this call with a fresh one carrying the same scopes. WARNING: this IRREVERSIBLY deactivates the current key the moment it succeeds — every other caller using that key stops working immediately, so do not call it on a shared credential. The new key is returned exactly once in the response and cannot be retrieved again. A delegated child key cannot rotate itself. Requires API-key authentication. |
| mint_child_api_key | Mint a short-lived child API key for a delegated subtask. The child can never exceed the calling key: requested scopes must be a subset of what the caller holds, its expiry is capped by the caller's, and it may never itself mint keys (delegation is exactly one level deep). A caller whose own scopes were never recorded explicitly cannot delegate at all until it rotates first. The child key is returned exactly once in the response and cannot be retrieved again; revoking the parent immediately revokes the child. Requires the keys:mint scope. |
| list_outreach_campaigns | List the authenticated user's outreach campaigns with live lead / sent / reply counts. Requires authentication. |
| search_outreach_leads | Search the user's outreach leads — filter by a text query (company / contact / email), pipeline status, and/or campaign. Returns company, contact, status, pitch, and reply info. Requires authentication. |
| outreach_analytics | Outreach summary for the user — total leads, sent, replied, and reply rate. Requires authentication. |
| score_locus_case | Score an anonymized adult mental-health / addiction case against LOCUS. Convenes the LOCUS Assessment Panel (psychiatrist, addiction specialist, clinical social worker, utilization reviewer, peer specialist, safety officer); each reviewer independently rates all six LOCUS dimensions, then a DETERMINISTIC engine aggregates the ratings and applies the Determination Grid and the inviolable override floors IN CODE (safety floors like Risk-of-Harm=4 → Level 5 cannot be reasoned away). Returns the recommended Level of Care with a full audit trail. Adults only (CALOCUS/CASII covers child/adolescent); use ONLY anonymized cases. Starts asynchronously by default; poll get_session. Requires authentication. |
| locus_determine_from_scores | Compute a LOCUS Level of Care from dimension ratings you already have. Deterministic — no LLM, instant: composite + Determination Grid + the inviolable override floors (e.g. Risk-of-Harm=4 → minimum Level 5), applied in code. Provide EITHER the seven flat D* ratings (1-5 each; Dimension IV splits into IV-A Stress / IV-B Support) OR a per-reviewer agent_scores map. Sending both is refused (422) rather than scored: agent_scores would win and your flat ratings would be discarded, override floors included. Requires authentication. |
| ask_legal | Request fixed legacy Themis source/citation records and attach a proposed Laws & Regulations agent/source/synthesis plan whose routes are planned_not_run. Check the law uses fixed legacy provider-record retrieval, but returned records do not prove exact citation identity, relevance, authority, or corpus coverage. Generated research requires authentication, exact THEMIS_NEUTRAL_SCHEMA_VERSION=themis_neutral_research/v1, the operator gate, a verified authenticated corpus manifest, and an atomic spend reservation. The manifest verifier is not available yet, so generated mode remains disabled. Provider assertions/gaps are enum codes linked to evidence whose quote-fidelity state is provider-reported and rendered with fixed server text; any legacy verdict or answer prose is withheld. Returns legal_evidence_graph/v1. Automated research only, not legal advice. Requires legal:run. |
| campaign_pipeline_stats | Pipeline-stage breakdown for one outreach campaign — total / sent / active / won / lost, reply rate, and per-stage lead counts. Requires authentication. |
| list_campaign_triggers | List automation triggers (auto-reply rules, status updates, notifications) configured for an outreach campaign. Requires authentication. |
| list_campaign_replies | List sent / received emails for an outreach campaign, newest first — subject, body preview, and reply classification. Requires authentication. |
| create_outreach_campaign | Create a new outreach campaign owned by the authenticated user. Returns the new campaign id. Reversible (campaigns can be edited/deleted in the dashboard). Requires authentication. |
| add_outreach_lead | Add a single lead to the authenticated user's outreach book. Returns the new lead id. Provenance is stamped source="agent" server-side. Does not send any email. Requires authentication. |
| assign_leads_to_campaign | Assign one or more of the user's existing leads to one of the user's campaigns (both must be owned by the caller). Returns how many were assigned. Idempotent. Requires authentication. |
| update_outreach_lead_status | Advance a lead through the pipeline — set its status / pipeline stage and, optionally, a recorded reply, pitch, or notes. The core pipeline-drive action. Idempotent (setting the same status twice is a no-op). Requires authentication. |
| draft_outreach_email | Save a draft email subject and/or body onto one of the user's leads. This only stores the draft for human review — it does NOT send anything. A human sends from the dashboard. Requires authentication. |
| create_outreach_pitch | Create a reusable pitch template (subject + body templates, optional drafting prompt) owned by the user. Provenance is stamped source="agent" server-side. Returns the new pitch id. Requires authentication. |
| list_deals | List the caller's sales deals (newest first) together with a weighted pipeline forecast (open/weighted/won totals). Optionally filter by stage or lead_id. Deals are private to the API-key owner. |
| get_deal | Fetch one of the caller's deals by id with its full activity timeline (notes, calls, meetings, emails, tasks; newest first). |
| list_sales_tasks | List the caller's open sales tasks (activities of type 'task' not yet done), bucketed overdue / today / upcoming and ordered most-urgent-first with per-bucket counts. Close one with complete_sales_task. |
| get_business_suite_overview | What needs attention across my business? Return the same private, counts-only review as the browser: Sales tasks, Marketing approvals, Accounting runs and Consulting work, with deterministic next actions. Legal is a governed entry point, not a verified review queue; generated Legal advice is disabled. No raw documents, amounts, contacts or private text are returned. Never sends, approves, pays or calls a provider. Requires the explicit business_suite:read scope, absent from default and legacy keys. Returns the canonical snapshot as JSON and structured data. |
| get_sales_analytics | Sales pipeline analytics for the caller: probability-weighted forecast, per-stage $ rollup, win rate (won / decided, by count and by value), average sales-cycle days over won deals, and open-deal aging with a stale count. Read-only; computed from the caller's own deals. Requires authentication. |
| get_sales_recommendations | The caller's prioritized next-best sales actions: interested leads to convert, open deals gone stale enough to need a follow-up, and overdue tasks — each with a rationale and the exact governed tool to run next (convert_lead_to_deal / log_deal_activity / complete_sales_task) plus its arguments. Read-only; ranks the caller's own CRM data (overdue > convert > follow-up). Nothing is executed or sent — approve an item by calling the named write tool. Requires authentication. |
| get_deal_health | Health scores for the caller's OPEN deals: each scored 0-100 on how likely it is to be slipping (healthy >= 70 / watch / at_risk), worst first, with plain-language reasons — days since last logged contact, whether it is past its own expected close date, and how its age compares to the caller's average won-deal cycle. Includes per-band counts, an average score, and the reference cycle. Read-only and deterministic over the caller's own deals + activities; nothing is executed or sent. Pair with get_sales_recommendations to act on what is slipping. Requires authentication. |
| create_deal | Create a sales deal owned by the caller. Provide a title, or a lead_id to inherit the lead's company/name as the title. The deal appears live on the owner's Sales board. No deletes are exposed over MCP. |
| update_deal | Update one of the caller's deals — most commonly to ADVANCE its stage (e.g. discovery -> proposal). Moving to closed_won/closed_lost stamps the close date; reopening to an open stage clears it. Only the fields you pass change; deal_id is required. |
| convert_lead_to_deal | Convert an owned outreach lead into a deal and advance the lead into the deal stage of the funnel (never downgrading an already-closed lead). Optional deal fields mirror create_deal; lead_id is required. |
| log_deal_activity | Log an activity on a deal (or lead) — a note/call/meeting/email, or a follow-up task with a due date. Pass deal_id and/or lead_id (each must be owned by the caller). type defaults to 'note'; for a task set type='task' and a due_date (YYYY-MM-DD). Tasks appear in list_sales_tasks. |
| complete_sales_task | Mark one of the caller's sales tasks/activities done (or reopen it with done=false). Stamps or clears the completion time; activity_id is required. |
| list_invoices | List the caller's invoices (newest first), optionally filtered by status (draft/sent/paid/void). Each entry includes its computed overdue flag. |
| get_invoice | Fetch one of the caller's invoices by id, including its line items. |
| get_invoice_aging | Summarize the caller's unpaid invoices by how long they have been past due. This owner-scoped, read-only report is computed live; an invoice is overdue only when its status is sent and its due_date is before today. Balances are separated by currency, with no conversion or combined money total. |
| create_invoice | Create a draft invoice owned by the caller for a client, optionally seeded with line items and linked to an existing deal. Totals (subtotal/tax/total) are computed from the line items and tax_rate. The invoice starts in status 'draft' — call send_invoice to mark it sent. |
| update_invoice | Update a draft or sent invoice's header fields (client, terms, due date, tax rate, notes). Only the fields you pass change. Blocked once paid/void — void and re-issue instead. invoice_id is required. |
| add_invoice_line_item | Add a billable line item to a draft or sent invoice; totals are recomputed automatically. Blocked once paid/void. invoice_id is required. |
| send_invoice | Mark a draft invoice as sent (stamps issued_at). Re-sending an already-sent invoice is a harmless no-op. invoice_id is required. |
| mark_invoice_paid | Mark a draft or sent invoice as paid (stamps paid_at). Cannot be applied to a void invoice. invoice_id is required. |
| void_invoice | Void a draft or sent invoice. Cannot be applied to an already-paid invoice (issue a credit note instead). invoice_id is required. No delete tool is exposed over MCP — void is the terminal non-destructive state. |
| convert_deal_to_invoice | Create a draft invoice from one of the caller's deals: one line item for the deal's amount, client_name defaulted from the deal's company/title. deal_id is required. |
| create_accounting_run | Analyze supported textual accounting records with the pinned deterministic Writeoff engine and create a private, encrypted audit run owned by the caller. Accepted filenames end in .txt, .md, .text, .eml, .csv, .ofx, .qfx, or .qif. Returns estimates for review only: it does not file taxes, move money, or send data to an external accounting service. Send prepared rows, not raw document text. Pasting the output of a PDF-to-text converter does not fail loudly. Read verbatim as .txt, .text or .md, the whole paste collapses to exactly ONE entry, and that entry's figure can be a real number from the page that is not a charge at all: a statement with no footer total books its running BALANCE as spend. Stripping the page headers and the balance column does not change that; a genuinely empty paste raises instead. Convert each document into canonical CSV first: a Date,Description,Amount header, then one row per economic event. The Description column carrying the merchant name alone is the cleanest form, and prose is not merely untidy -- it CHANGES the answer. The matcher runs over the whole description against a vocabulary that holds ordinary words as well as vendor names: paper, printer, ink, notebook, legal, consulting, subscription, hosting and domain each classify alone with no vendor present, and contact lens or reading glasses classify as a MEDICAL deduction. A memo line saying what was bought can therefore create a deduction the vendor name alone would not. Send the merchant, not a description of the purchase. Not every phrase matches -- weekly grocery run, haircut and banana all stay unclassified -- but plain English is not inert. Classification is separately gated by context_text, but that gate is PARTIAL, not an off switch. Leaving it empty suppresses only the rules that need a business to exist: business, meals, vehicle and home-office. The personal rules stay live whatever you send -- a bare contact lens still classifies with context_text empty, under the category name fsa_hsa rather than anything called medical, and donation and tuition classify the same way. An empty context is NOT a way to stop deductions being proposed. The business gate is a bare substring test, not a reading of what you wrote: it opens on a keyword anywhere in the text, so I do not own a business and This is a personal return, not a business each switch the business rules ON rather than off. Negation is not detected. Suppress those rules with an EMPTY context, never with a denial. Booking is independent of all of it -- a row books its amount whether or not it classifies, and classification decides only whether the row becomes a deduction candidate. State the direction of every row. A .csv is routed as a statement, so each row has to say whether money went out or came in. Send exactly Date,Description,Amount,Type -- the amount column named Amount, and a Type of Purchase for money out or Refund for money in. Separate the fields with COMMAS. The delimiter is sniffed from the header row, and a semicolon additionally switches the amount parser to the European convention where the comma is the decimal point and the dot is a thousands separator. In a semicolon-delimited file an ordinary -20.00 is therefore read as 2000.00 and -1234.56 as 123456.00 -- silently, with no error, a hundred times the real figure. Tab and pipe keep the dot decimal. Send these four charge columns, each once; the optional duplicate-review metadata columns described below are also supported. Order among the four does not matter -- all 24 arrangements of Date,Description,Amount,Type measured identical -- and duplicate CSV headers are rejected before any transactions are returned. Names are compared after Unicode NFKC normalization, trimming outer whitespace and casefolding, including unknown and repeated blank headings. No copy is chosen, even when the repeated columns contain equal values. A literal Amount protects the figures from same-direction ancillary debit- or credit-looking columns. A literal Amount beats one or more same-direction ancillary columns in either order, whether those cells are populated or blank. Those ancillary values are not silently substituted. Opposite-direction ancillary columns together form a complete pair and fail closed beside Amount. Schema arbitration happens before row direction; Type cannot rescue an ambiguous schema. After arbitration selects one money representation, row direction uses a recognized Type first. Only when Type is absent or unrecognized does a trailing CR or DR marker decide. Only when both are absent does the selected money heading or sign decide. Thus Purchase and Refund outrank a conflicting marker on an otherwise unambiguous schema; notably, Payment is unrecognized and preserves the fallback. Conflicting money representations fail closed instead of being chosen by header order. Distinct equal-ranked amount aliases, multiple equally ranked same-direction money headers without a literal Amount, an amount alias beside a separate debit/credit representation, a literal Amount beside a complete pair, and a third amount candidate beside a pair each produce a named ambiguous-money error. A complete Debit/Credit pair remains supported, but a row with both pair cells nonzero fails closed. On a row with neither recognized Type nor CR/DR marker, a negative debit is a reversal and remains credit, while a negative credit is never promoted to spend. Within one role vocabulary, exact matches still beat partial matches. Column roles remain isolated. Description prefers an ordinary non-role heading. If none exists, exactly one releasable semantic-directional heading such as Charge Description or Payment Memo may serve, but only when an independent money representation survives without it. Outside that semantic-directional exception, suppressed lower-tier money, date and type candidates remain reserved and cannot become Description merely because a stronger sibling won their original role. Structural composites such as Amount Details, Transaction Type Description and Debit Details, or multiple competing semantic candidates, cannot serve as Description; when only those remain, the file fails closed rather than poaching merchant text. A duplicate-heading source contributes no transactions or amounts and carries a correction in catalog.errors. Inspect those errors even when other valid files let the batch complete. Check the original export and give each column a unique name before submitting a corrected file; saved historical results are not rewritten. Type is matched against a fixed vocabulary, not read as free text. purchase, debit, charge, withdrawal and dr mean money out; refund, credit, deposit, return and cr mean money in. All eight spelled-out words resolve in the plural as well, but the two abbreviations do not: drs and crs are unrecognised and fall through to the amount CELL, so a crs row written negative books as SPEND, not as money in. Anything else -- notably Payment, money in on a card but out on a checking account -- counts as unstated, and the direction then falls to the amount CELL rather than to the sign alone. Write the amount as a plain signed number, with nothing else in the cell. A trailing DR or CR and accounting parentheses are not decoration, and what they do depends on the column holding them and on the parser the CONTENT selected -- NOT on the file suffix. Two separate things happen to such a cell, and BOTH are confined to the statement path -- the fan-out described under Routing below. First, in every statement format and every column, the figure is given a sign: parentheses negate, DR negates, and CR does nothing at all. DR is a SIGN; CR is only a label. On the single-receipt path no marker is a sign at all: a total written 20.00 books 20.00, and that same total written 20.00 DR, (20.00) or 20.00 CR books 0.00 -- there the marker makes the amount UNREADABLE rather than negative. Second, when the content parses as CSV and schema arbitration has selected one money representation, row direction reads recognized Type first and then any marker on the selected money cell. A marker never overrides recognized Type; with Type absent or unrecognized it outranks the selected heading or sign. On a bare Date,Description,Amount file, 20.00 DR and (20.00) are both money OUT, and even -20.00 CR is money IN. A recognized Purchase or Refund still outranks either marker. A cell carrying no marker falls back to the SIGN, under the bank convention where money out is NEGATIVE, so a plain POSITIVE amount reads as money coming in and is dropped as non-deductible. On three rows totalling 137.19: written plain and positive they record nothing, and those same positives written 20.00 DR or (20.00) record all three. This CSV marker reading applies to the selected single money column or the selected nonzero cell of a complete pair. With Type absent or unrecognized, 20.00 CR is credit and 20.00 DR is debit under Amount, Charges or Payments. With neither recognized Type nor marker, heading/sign fallback remains: positive Charges is spend, negative Charges is a reversal, and a lone Payments column is credit. Beside literal Amount, even a blank Payments column is ancillary and cannot void honest figures. A money column whose name is in NEITHER vocabulary is a third way to record nothing. Purchases, Spend, Cost and Total were each measured doing it -- they are examples, not a list to check yours against -- and a file whose only figures sit under such a name returns zero rows at BOTH signs, with no error. Recognition is by name against a closed list, so the remedy is not a clearer word of your own but the four columns named at the top of this contract. Native OFX and QIF content behaves like a plain AMOUNT column and NOT like a money-out column, on all six markers, under .ofx, .qfx and .qif alike: the sign left by the first step decides, so -20.00 CR is money OUT, and so is (20.00 CR); 20.00 DR is money OUT because DR negated it; and a plain positive 20.00 is money IN and dropped. TRNTYPE is never consulted -- DEBIT with a positive TRNAMT still drops. Because the parser follows the CONTENT, a .qfx holding canonical CSV runs the CSV rules above instead, markers and all. Dropped rows are silent, and nothing in the result marks a row as dropped. A run does fail when it analyzed nothing at all across the whole submission, and separately on transport, input and engine errors -- but no failure mode reports a PARTIAL loss. If even one row anywhere survives, the run completes and the rest vanish with no notice, so a completed run is NOT evidence every row was read. A three-row file with one negative amount and two positive ones returns one row, no error, and a total indistinguishable from an honest one. The one loss that IS named is a document that contributed nothing at all. When a submitted file appears in no catalog channel -- no item, no unclassified row, no notice, no error -- and it held at least two non-empty lines, catalog.notices carries a source_left_no_trace entry naming that file. That covers whole-file loss: a statement whose rows use a different delimiter than its header, and a body of unreadable bytes under a valid header, both otherwise return the same empty success as a file that genuinely held nothing. A header-only export stays silent by design -- it has no body to lose. This does not narrow the PARTIAL case above: a file that recorded even one row counts as read, so rows dropped beside it remain silent. Routing is by filename suffix, not by content, against a CLOSED allowlist -- and the allowlist is assembled from two constants that DISAGREE. SUPPORTED_SUFFIXES in web/accounting_engine.py admits eight: .csv, .ofx, .qfx, .qif, .txt, .text, .md and .eml. STATEMENT_SUFFIXES in writeoff/batch.py names the five that fan out into one row per line, and one of those five is .xlsx, which the engine refuses before any parser sees it. What fans out is the INTERSECTION -- .csv, .ofx, .qfx and .qif. The other four -- .txt, .text, .md and .eml -- are read as a single receipt. Of those, .txt, .text and .md are read VERBATIM and behave identically to each other; .eml is NOT one of them. An .eml is parsed as an email FIRST -- headers dropped, transfer-encoding decoded, an HTML body flattened into lines at its block tags -- so every rule below applies to THOSE lines and not to the file's, and the same bytes can book a different figure, name a different merchant, or record a load error and contribute no entry at all. Statement rows sent that way collapse to ONE entry, and what that entry books turns on the PRICE pattern below: a file that LOADS and in which NO line ends in a price books 0.00, silently and with no error. An .eml with no extractable body -- an attachment-only mail -- never reaches that stage: it contributes NO entry and records an EmailIngestError in the run's errors, while still being listed among its sources. Its merchant is NOT the file's first line: it is the first line that neither ends in a price nor is a bare date, so a file led by a CSV header books that header text as the merchant, and a file whose every line ends in a PRICE books an EMPTY merchant. Ending in a bare number is not enough: a Closing balance 900 line is itself booked as the merchant. The winning figure is chosen by matching against the WHOLE lower-cased LINE, description included, so a purchase from TOTAL WINE AND MORE reads as the file's total. A label is only ever read on a line that ENDS in a price, and a price means EXACTLY TWO DECIMAL PLACES: a whole-dollar Total 137 is not a price, and neither are 137.1, 137.190 or a trailing 137. -- none of their labels are ever read. The pattern is anchored at the END only, so what stands in FRONT of the figure is unrestricted: Total USD 137.19 and even Total about 137.19 both read as totals, and a leading dollar sign is merely one case of that. After the digits it admits an optional minus and at MOST ONE trailing letter, and those letters are UPPERCASE ONLY -- T, N, X, F, E or an asterisk, while a lowercase t, n, x, f or e leaves the line with no price at all. So a line reading Total 137.19 USD, Total 137.19 (USD) or Total 137.19 EA carries no price at all, its label is never read, and the file falls through to its largest amount -- and a payable line behaves the same way, so an Amount Due 137.19 USD is not a payable line either. That property, not the marker, is why a Total 137.19 CR is not read as a total: CR is two letters, so the line has no price. Otherwise any line containing total is a total line and the LAST one wins -- not the first, not the largest -- so that wine row REPLACES an honest footer standing above it. Excluded are subtotal and sub total, though the near-miss sub-total is not excluded and does win, and any total line also holding saving, save, discount, coupon or reward as a bare substring; that is the entire exclusion list in the pinned dependency today. Those exclusions are the dangerous half: an excluded line is read as NO total rather than as a smaller one, and the search moves past it to the payable stage below, with no zero total present, and only then to the largest amount. Rows of 4812.00 and 12.34 under a Total Rewards Earned of 42.10 book 4812.00 -- a hundredfold over-book off a line the file itself labels a total. Give that same file an Amount Due of 42.10 and the payable stage rescues it. A ZERO is not an exclusion and does not chain that way. A total line reading 0.00 WINS its stage and sets the total to zero, and a zero total SUPPRESSES the payable stage, so the file drops to its LARGEST amount in EITHER order: an Amount Due of 42.10 standing above or below a Total of 0.00 does not rescue it. A paid-in-full invoice reading Total Due 0.00 books its largest line item. Last-wins covers zeros too, so a Total of 0.00 below a real Total of 42.10 WIPES it. The payable stage does not behave that way -- it takes the last payable line whose figure is NON-ZERO, so a trailing Amount Due of 0.00 leaves an earlier Amount Due of 42.10 standing. With no total line surviving, a line reading amount due, balance due, amount payable or please pay is used instead, last-wins again and only when its figure is non-zero. Those four are matched as LITERAL text, so the doubled-space spellings Amount Due and Balance Due, and likewise Amount-Due and AmountDue, are NOT payable lines at all; a file whose only payable label is spelled one of those ways skips to its LARGEST amount, which is the direction that over-books. A trailing colon in Amount Due: still matches. Failing every stage, the largest amount anywhere in the file. One hazard, though, is NOT confined to that last stage. The price read is END-anchored on the LINE, not on a column, so it reaches back across commas and swallows text belonging to the field BEFORE it. It takes the line's final cents pair and walks LEFT across each comma-separated group of EXACTLY three digits, then swallows up to three trailing digits of whatever text precedes the first such comma -- an amount's cents, a check or invoice number, a card last-4, a units or store-number column; money or not, decimal point or not. A group of one, two, or four-or-more digits terminates the walk. So a row of -12.34,887.66 yields 34,887.66 -- a figure that appears in NO cell of the document, even though those characters occur across the comma between two cells in the raw text. That splicing happens while the price is being READ, which is before any label is tested, so a spliced figure is what a total line or a payable line CARRIES INTO its own stage: a footer reading Total,-60.34,887.66 books 34,887.66 rather than 60.34, and an Amount Due,-12.34,887.66 books 34,887.66 the same way. Holding an explicit Total line is therefore NOT a way out of this. The field on the LEFT need not be money and need not carry a decimal point: a check register whose amount is the LAST column books 140,732.19 from a row reading 2029-06-02,8140,732.19, and an Amount Due,INV 4522,887.66 books 522,887.66. There is no ceiling on the RIGHT either -- -31.20 beside 412,880.55 books 20,412,880.55, because 412 and 880 are each groups of exactly three. What stops the walk is GROUP WIDTH alone, so a right-hand 1000.00, 4,887.66 or 12,345.67 is read whole, and the total line then books THAT figure rather than its own. A spliced figure is usually in no cell of the document, but not always: when the digits swallowed are all ZEROS -- an amount's 00 cents, or a round 1000 -- it lands exactly on the right-hand figure, so agreeing with a real cell is not evidence of a clean read. No error is raised in any of these cases. Send that file as .csv instead, where the splice does not occur. Rows of 200.00 and 12.34 under a Total of 137.19 book 137.19; delete that total line and the same file books 200.00. The winning line's own minus sign is DISCARDED at either stage, so a Total of -137.19 and an Amount Due of -137.19 both book 137.19. The entry books 0.00 whenever NO line in the file ends in a price, and a DR, CR or parenthesis marker on every amount is only ONE way to reach that state: whole-dollar amounts, a trailing currency code and a trailing period each book 0.00 with no marker anywhere in the file. Where a marker IS the cause, leaving the winning line unmarked makes it book in full, whether the winner is a total or a payable line. Every rule in this paragraph is measured on the LINES the loader hands the parser: for .txt, .text and .md those are the file's own lines, and for .eml they are the extracted body's. Anything outside the eight is REFUSED outright with an unsupported-file-type error rather than silently mis-read, so .pdf and .xlsx never reach the parser. Because the second set lives in a pinned dependency, treat the four as measured today, not as a promise. That split is by suffix; WITHIN those four the parser is chosen by CONTENT, so a .qfx holding canonical CSV runs the CSV column contract above in full. Preserve each verified source row. Exact equal amounts for the same merchant (case/whitespace normalized) within 14 days inclusive are candidate pairs. Missing dates and explicit shared Event ID also require review. Disputed contributions are excluded from all counted totals and shown as pending review; pending source-occurrence amounts are not unique-payment estimates. Raw documents and references remain evidence. Different invoice/order/card references alone never prove separate payments. Use duplicate_resolutions with exact source/row locators, a review_note and decision distinct_payments or same_payment; the latter requires an explicit counted_source and conflicting amounts/itemization stay pending. Create a new run with replacement choices to recalculate, or [] to reverse them; historical runs stay unchanged. At most 500 choices, 500 sources each, and 10,000 expanded reviewed pairs. Optional CSV columns Event ID, Payment Reference, Payment Review and Payment Review Note preserve evidence. Only explicit reviewed distinct rows with Payment Review=distinct_payment, different Event IDs and nonempty notes restore separate contributions without a resolution list. Scope each run to one entity and one tax year, and name it that way -- totals spanning entities or years match no filing. Book each event on the date money moved (cash basis) unless the entity files accrual, and never mix bases in one run. Submit runs one at a time: one ingest is active per owner at a time. The returned total counts classified spend only. Charges whose merchant matches no deterministic rule are excluded from it, so do not present it as total spend. Use preview_accounting_ticket_sync to turn those residuals into reviewable work. Never submit a figure you cannot find verbatim in the source document, and never adjust a merchant name to make a row classify: Description is what the engine matches on, so renaming an unrecognised vendor to a recognised one raises the deduction while every figure stays verbatim. An unclassified row is the engine declining to assert a rule it does not have. Leave it, and report the count. |
| list_accounting_runs | List only the caller's private Accounting-run audit metadata, newest first. Raw source text and analysis results are intentionally omitted. |
| get_accounting_run | Get one private Accounting run owned by the caller, including its deterministic estimates, disclaimer, engine revision, and audit timestamps. Source input stays omitted unless include_input=true is explicitly requested. |
| delete_accounting_run | Permanently delete one private Accounting audit run owned by the caller. This is the only Accounting MCP deletion; it never files taxes, moves money, or contacts an external service. |
| preview_accounting_ticket_sync | Preview, without writing, how one completed owned Accounting run's WRITEOff work plan would map to stable Meta Council tickets. Reports creates, updates, unchanged tickets, human-edited generated fields that will be preserved, and identity conflicts. Requires both accounting:read and tickets:read; tickets:write also satisfies the ticket-read grant. Preview before committing because tickets are not field-encrypted; work-plan text may reproduce source-derived snippets or parser details, and after commit tickets:read can read it without accounting:read. Check the item count before committing. Statement-derived runs currently plan one ticket per unclassified transaction line, so a several-hundred-row statement plans several hundred tickets. When the count is large, report it and confirm with the owner rather than committing a board-flooding sync. |
| sync_accounting_tickets | Commit a previously previewed completed Accounting run's WRITEOff work plan to stable owner-private Meta Council tickets. The expected_plan_hash from the preview is required, retries are idempotent, stable references prevent duplicate tickets across reprocessed runs, and human edits to generated content fields are preserved. Status, assignee, hierarchy, and completion are not reset. Requires both accounting:read and tickets:write. This creates or updates only in-platform planning tickets; they are not field-encrypted and become readable through tickets:read without accounting:read. It never files, pays, sends, or publishes anything. |
| ticket_plan | Preview or idempotently commit an owner-private ticket tree generated from a planning prompt or supplied structured plan. Requires tickets:write. A preview without a supplied plan invokes the planner model and additionally requires councils:run; supplied-plan preview and commit are provider-free and need only tickets:write. Preview is non-mutating and returns the normalized commit payload, an exact preview_token, and deterministic predicted ticket IDs as JSON. To commit, combine that commit_payload with mode=commit, the caller-held idempotency_key, and that exact preview_token; the response returns the same IDs plus replay status. A caller-stable visible-ASCII idempotency_key is always required. This creates only in-platform planning tickets; it never runs them, sends, publishes, deletes, or changes an external provider. |
| ticket_batch_create | Preview or idempotently commit one atomic owner-private batch of structured tickets and nested subtickets. Preview is non-mutating and returns the normalized commit payload, exact preview_token, and deterministic predicted IDs as JSON. Commit requires that exact token: combine commit_payload with mode=commit, the caller-held idempotency_key, and preview_token. It returns the same IDs plus replay status. A caller-stable visible-ASCII idempotency_key is always required. MCP authorship is stamped by the server; callers cannot spoof it. Nested subtickets are bounded and validated by the ticket handler before any write. This never executes, sends, publishes, deletes, or mutates an external provider. |
| ticket_list | Page through the caller's tickets on the Meta Council board. The opaque next_cursor is owner- and filter-bound; keep every filter unchanged on the next call. Immutable cursor ordering keeps a stable full-board traversal exact while returned tickets are edited or reordered. Filter by status, assignee, action_type, priority, parent_id (a ticket UUID, or 'none' for root tickets only), or free-text q over title/description. With recursive=true, parent_id must be an owned UUID and all descendants (not the anchor) are returned as one flat, cycle-safe traversal. Session/workflow links are opaque metadata filters, not access grants. Start here to find work, then use ticket_get for detail and ticket_claim to take a ticket. output_format=json returns a minimized machine-readable page inside text, including labels and stored update timestamps, for read-only board mirrors. It is not a full ticket backup. |
| ticket_get | Full detail for one owned ticket: description, acceptance criteria, subticket progress, and the recent activity/comment trail. ticket_id is required. |
| get_ticket_projects | Projects overview of the caller's board: one entry per top-level epic with recursive descendant progress (done %, counts by status, blocked, points, sub-epics, last activity), plus a triage strip of unrouted work (backlog + unassigned root tickets). Read-only, recomputed live. Use this before planning a session to see initiative health at a glance; use ticket_list/ticket_get to drill into any id it returns. |
| get_workspace_metrics | One cross-domain rollup of the caller's own workspace: council sessions (total, last 30 days, by status, tokens), feedback ratings, tickets (open/done), deals (count by stage, open pipeline value) and invoices (count by status, outstanding total, overdue count). Read-only, recomputed live, and scoped to the caller — it never aggregates across accounts. Amounts are labelled and grouped by recorded currency, never converted or added across currencies; unavailable currency or amount coverage is stated explicitly. Use it for a single 'how is this workspace doing' answer instead of calling the per-domain analytics tools one by one; use those (get_sales_analytics, get_ticket_projects, list_invoices) to drill into whatever this surfaces. Counts of unrecognised stages or statuses are reported under 'unknown' rather than dropped, so each breakdown sums to its own total. Note that open + done need not equal the ticket total: cancelled tickets are neither. Invoice overdue status is computed at read time by comparing due dates against now, not stored on the record, so it is current as of this call and an invoice due today does not yet count as overdue. |
| get_meta_council_platform_metrics | OPERATOR ONLY: cross-owner analytics for the META COUNCIL PLATFORM itself — every account added together, NOT the caller's workspace (use get_workspace_metrics for that). Requires both the platform:admin scope AND an ADMIN_EMAILS operator account; everyone else gets a permission error. Returns content-free aggregates only: account counts, 30-day active owners, session counts by status and token totals, ticket open/done, deal pipeline and outstanding invoice amounts separated by recorded currency without FX conversion and with missing coverage stated, overdue invoice counts, feedback backlog, per-pillar adoption, the busiest panel slugs, and a per-account activity table (email and counts). It never returns query text, answers, feedback bodies, deal or invoice detail, or any other text a user typed — the aggregate reports how much, never what about. Note that open + done need not equal the ticket total: cancelled tickets are neither. |
| get_site_analytics | OPERATOR ONLY: visitor traffic for the META COUNCIL PLATFORM's own website — every visitor added together, NOT the caller's workspace (use get_workspace_metrics for that) and NOT the business rollup (use get_meta_council_platform_metrics for that). Requires both the platform:admin scope AND an ADMIN_EMAILS operator account; everyone else gets a permission error naming which of the two requirements failed. Covers a window of whole UTC days ending today, set by days. Returns total views, unique visitors, the authenticated/anonymous split, the busiest public paths, the busiest in-app sections, top referrers and a per-day series, plus the window it covers so the caller need not track what it asked for. Page paths and in-app sections are counted separately and are not comparable to one another. It never returns per-visitor rows, email addresses, IP hashes, user agents, session identifiers or query text — the rollup reports how many, never who. |
| get_ticket_burndown | Daily burndown series replayed from the append-only ticket ledger: per-day scope (existing, non-cancelled), remaining, done levels plus added/completed flows, with a summary. Optionally scope to one epic's current subtree via root_id. Backfill (legacy_snapshot) rows seed state but never count as additions. days clamps to 7-180 (default 30). Optionally pass target_date to overlay a straight-line plan: an ideal series descending from the window-start remaining to zero on that date, for reading actual against plan. |
| list_saved_views | List saved ticket views you can open: the stored filter, column, grouping and sort combinations saved by you, plus any shared with a team you belong to. Shared views appear under the active and all scopes and never once archived, because an owner's archive is their own working state. Returns the same payload the web app's saved-view list returns. Requires authentication and the tickets:read scope. |
| get_saved_view | Fetch one saved ticket view's definition by id: its name, filters, columns, grouping, sort and whether it is shared. This returns the view itself, not the tickets it selects -- use execute_saved_view to run it. A view you cannot read is reported as not found rather than as forbidden, so this cannot be used to discover that someone else's view exists. Requires authentication and the tickets:read scope. |
| execute_saved_view | Run a saved ticket view's stored filters against live tickets and return the matching tickets, the view that produced them, and group counts when the view groups. Results are always YOUR tickets: a shared view shares the question, not the answers, so opening a teammate's view runs it against your own work and never reveals theirs. An archived view is refused rather than executed. Returns the same payload the web app receives. Requires authentication and the tickets:read scope. |
| list_portfolio_snapshots | List your own portfolio snapshots, newest first: the immutable rollups of how your ticket portfolio stood at the moments you captured them. Each carries its label, schema version, stored payload, content hash, recording actor and creation time. Snapshots are owner-private and never span accounts. Returns the same payload the web app's snapshot list returns. Requires authentication and the tickets:read scope. |
| get_portfolio_snapshot | Fetch one portfolio snapshot by id: its label, schema version, the exact stored rollup payload, its content hash, who recorded it and when. The payload is served from storage and is never recomputed from live tickets, so a snapshot read today reports what was true when it was taken. A snapshot you do not own is reported as not found rather than as forbidden. Requires authentication and the tickets:read scope. |
| diff_portfolio_snapshots | Compare two of your own portfolio snapshots and return the per-dimension movement between them. snapshot_id is always the 'to' side; against is the 'from' baseline, defaulting to your newest snapshot strictly older than it -- the 'what changed since last time' read, which is reported as not found when this is your first snapshot. Comparing an older snapshot against a newer one is allowed and disclosed rather than silently confusing: the result says whether the pair is chronological and the elapsed days go negative. Both sides are immutable, so a diff is reproducible indefinitely. Requires authentication and the tickets:read scope. |
| list_portfolio_snapshot_notes | Read the decision and comment thread recorded against one of your portfolio snapshots, oldest first. Each note carries the snapshot content hash it was written against plus binding_intact comparing that hash to the snapshot's hash now, so a note can be read as evidence of what was actually on the screen when it was written. Counts are over the notes actually returned, so a truncated page never reports a total it did not show. This reads the thread; it does not approve, reject or block anything. Requires authentication and the tickets:read scope. |
| create_saved_view | Save a named ticket filter, column, grouping and sort bundle so you can reopen it later or share it with a team. Saves the QUESTION, not an answer: the view is re-run against live tickets every time it is opened, so it never goes stale. The new view is private until you share it. Same endpoint the web app's save button uses. Requires authentication and the tickets:write scope. |
| update_saved_view | Revise one of YOUR OWN saved views in place, keeping its id and any existing team share. Only the fields you send change; the rest are left exactly as they were, so this is safe to call without re-reading the whole definition first. Sharing a view does not let its readers revise it -- only the owner may. Requires authentication and the tickets:write scope. |
| archive_saved_view | Retire one of YOUR OWN saved views without destroying it: it drops out of the default listing and execute_saved_view refuses it, but its definition is kept and restore_saved_view brings it back unchanged. Prefer this over delete_saved_view. Idempotent -- archiving twice keeps the original archive time. Archiving is your own working state and does not hide the view from a team it is shared with. Requires authentication and the tickets:write scope. |
| restore_saved_view | Return one of YOUR OWN archived saved views to the working set with its definition intact. Idempotent -- restoring a view that is already active is a no-op rather than an error. Use list_saved_views with scope=archived to find restorable views. Requires authentication and the tickets:write scope. |
| share_saved_view | Share one of YOUR OWN saved views with a single team you belong to. Members may read, list and execute it; only you may revise, re-share or delete it. Executing a shared view runs it against the READER's own tickets, so this shares the question and never the answers. Re-sharing to a different team MOVES the view rather than adding a second grant -- a view is readable by at most one team. Note: team ids are not currently discoverable from this MCP surface, so supply one you already hold. Requires authentication and the tickets:write scope. |
| unshare_saved_view | Make one of YOUR OWN shared saved views private again. Idempotent. Revocation is immediate: team membership is resolved on every read and never copied into a grant, so this takes effect on each member's next request. Requires authentication and the tickets:write scope. |
| delete_saved_view | Permanently delete one of YOUR OWN saved views. UNRECOVERABLE -- there is no undo and the definition is not retained. archive_saved_view is the reversible retirement and is almost always what is wanted instead; use this only for a view that should actually be gone. Deleting a shared view removes it for the team too. Requires authentication and the tickets:write scope. |
| create_portfolio_snapshot | Capture an immutable rollup of YOUR OWN ticket portfolio as it stands right now -- totals by status and priority, fixed at this moment and never recomputed. A snapshot is the durable 'here is where we were' that a later diff_portfolio_snapshots measures movement against, so capture one before a review rather than after. Snapshots cannot be edited or deleted once taken. The snapshot records the credential that captured it, so one taken by an agent is attributable as such. Requires authentication and the tickets:write scope. |
| create_portfolio_snapshot_note | Record a decision or comment against one of YOUR OWN portfolio snapshots. Pass expected_content_hash -- the hash you saw when you read the snapshot -- and a mismatch is REFUSED rather than quietly filed against a different revision; that check is the point of this tool. Omitting it still records the note, it just does not assert which revision was on the screen. This writes down what was decided; it does not approve, reject or block anything, and nothing downstream gates on it. Requires authentication and the tickets:write scope. |
| ticket_validation_finalize | Finalize one immutable validation run for an exact owned ticket scenario revision. Requires the opt-in tickets:validate scope. The server derives owner and actor only from the authenticated API key. Evidence payloads are bounded opaque JSON objects; locator-looking strings are recorded but never opened, resolved, redirected, or fetched. Exact retries by the same credential actor return the original receipt; a changed actor or changed content under the same key conflicts. |
| ticket_scenario_create | Author one Given/When/Then scenario on an owned ticket. This is the specification an epic must carry before it can be completed: record a passing run against it with ticket_validation_finalize, then transition the epic. The server owns revision 1 and returns current_version and current_definition_hash — pass those exact values to ticket_validation_finalize. Authoring a scenario is not evidence; it states what must be proven, not that it was. |
| ticket_scenario_revise | Replace an owned scenario's content, compare-and-swapping on expected_version. Revisions are append-only: validation runs already recorded keep naming the exact version and hash they were proven against, so revising never rewrites past evidence — it does mean the epic needs a fresh passing run against the new version. A stale expected_version conflicts rather than overwriting a concurrent edit. |
| ticket_readiness | Explain why an owned ticket is not ready to close, using the same readiness check as the browser. Returns gated, ready, and the complete bounded blocker list with each code, reference_id and message. An unready ticket is a successful diagnostic, not a tool failure. This snapshot does not authorize completion or change tickets, scenarios, evidence or validation-attempt budgets. Requires authentication and tickets:read; tickets:write also permits reading. |
| ticket_scenario_list | List an owned ticket's scenarios with their exact current revisions, including the current_version and current_definition_hash that ticket_validation_finalize requires. Read-only. Use it before recording a run to confirm which revision is current, and to see why an epic still reports required_scenario_missing. |
| ticket_dependency_add | Record that an owned ticket cannot complete until another owned ticket is done. The edge is never erased: correct a mistake with ticket_dependency_remove, which marks the edge removed and appends to a ledger rather than deleting anything. Re-adding the same prerequisite reinstates that original edge instead of creating a second one, so an edge removed and restored says so. An edge stops blocking when the prerequisite reaches done, and only then — archiving the prerequisite does not release it. Only epics are gated, so an edge on a task is recorded and reported but never blocks its completion. Rejections carry a stable code: dependency_self (422), dependency_duplicate, dependency_cycle, dependency_graph_too_large (409), ticket_not_found (404). |
| ticket_dependency_remove | Stop an owned ticket from waiting on a prerequisite. Removing the last unsatisfied edge can allow this ticket to close, so this is a gate-opening action. Nothing is deleted: the edge is retained, marked with who removed it and when, and the removal is appended to a ledger that survives everything after it. This is not a way to say the prerequisite work is finished — that remains a separate statement about that ticket. Removal is allowed even when either ticket is archived. Rejections carry a stable code: dependency_not_found (404), dependency_already_removed (409). |
| ticket_dependency_list | List the prerequisites an owned ticket waits on. Read-only. Each edge reports satisfied, which is true only when the prerequisite is done, plus active/removed_at/removed_by_actor. Removed edges are omitted unless include_removed is true; they never gate completion, and are readable as history. |
| ticket_create | Create a ticket on the caller's board — optionally as a subticket via parent_id. Recommend degree of difficulty with effort (trivial|small|medium|large|epic) and the kind of work with action_type (strategy|implementation|research|validation|testing|coordination). The authenticated API-key UUID is stamped by the server as creator provenance; callers cannot supply or spoof it. Session/workflow ids are opaque metadata links only. New epics cannot be created already done. |
| ticket_update | Update fields on an owned ticket — status moves (each one is auto-logged to the ticket's audit trail), reassignment, priority/effort/action_type recommendations, title/description edits, safe reparenting, opaque session/workflow links, effort points, order, and user-owned external refs. Clear flags are explicit so stdio clients never lose JSON nulls. System-owned provider, Accounting, plan, and batch refs are preserved and cannot be spoofed. The authenticated API-key UUID is stamped on the audit trail. Epic completion and terminal reopen fail closed on this generic tool until an explicit lifecycle operation supplies retry/version evidence. |
| ticket_archive | Retire an owned ticket from the working set, or restore one. Archiving is a different axis from status: the ticket keeps whatever status it had, so a done ticket stays done and a blocked one comes back still blocked — never use status 'cancelled' to mean 'archived'. An archived ticket disappears from ticket_list, the stats, the buckets, the attention queue, and outbound JIRA/Linear sync, but keeps its id and URL so citations stay valid, and it refuses edits until restored. Cascades to the active subtree; restore brings back exactly what was archived alongside it. Nothing is destroyed — this is not a delete, and no delete tool is exposed. |
| ticket_comment | Add a comment to an owned ticket's activity trail. Use kind='progress' for work updates while a ticket is in progress. The authenticated API-key UUID is stamped as author; callers cannot supply or spoof it. Open the comment in plain language before any technical detail — these are read during escalations by people who were not part of the work. |
| ticket_claim | Claim a ticket to work on: sets assignee to your agent name and moves it to in_progress in one step (the move is audit-logged). Refuses if another agent already has it in progress unless force=true. Follow up with ticket_comment progress updates and finish via ticket_update status='done'. |
| submit_meta_council_feedback | Report feedback about the META COUNCIL PLATFORM itself — a noticeable performance failure, a lacking/missing capability, a bug, or a UX/docs gap in Meta Council's tools, councils, workflows, or MCP surface. Platform feedback only: do NOT use this for anything about your own business, customers, deals, invoices, or other data Meta Council manages on your behalf — use the relevant business tool for that instead (e.g. create_deal, create_invoice). Use it the moment a tool result, council run, or workflow falls short. Stored per-user and visible only to platform admins — the response returns an id + acknowledgement and submissions cannot be read back. Include machine context (tool name, session id, model) in the context object. |
| list_meta_council_feedback | ADMIN ONLY: list platform-feedback reports — feedback about Meta Council itself, never a tenant's own business data — across all users for triage (requires an ADMIN_EMAILS account; everyone else gets a permission error). Filter by status, category, or severity. |
| triage_meta_council_feedback | ADMIN ONLY: triage a platform-feedback report — set its status and/or admin notes (requires an ADMIN_EMAILS account; everyone else gets a permission error). Get the full feedback id from list_meta_council_feedback. |
| list_marketing_brands | List brand identities owned by the API-key account. |
| get_marketing_brand | Get one owned brand identity with its voice, value proposition, and guidelines. |
| list_marketing_audiences | List reusable marketing audiences owned by the API-key account. |
| get_marketing_audience | Get one owned audience definition with pain points and channels. |
| list_marketing_campaigns | List owned marketing campaigns. These are separate from Sales outreach campaigns. |
| get_marketing_campaign | Get one owned marketing campaign with its brand, audience, channels, and dates. |
| list_content_assets | List owned content assets and immutable revisions; content is never published by this tool. |
| get_content_asset | Get one owned content revision, including its review state and exact content hashes. |
| get_content_calendar | List scheduled owned content between optional ISO-8601 boundaries. Scheduling is planning metadata, not publishing. |
| create_marketing_brand | Create an owner-private brand identity. Requires marketing:agent_write. |
| update_marketing_brand | Update fields on one owned brand identity. No delete is exposed over MCP. |
| create_marketing_audience | Create an owner-private reusable audience definition. |
| update_marketing_audience | Update fields on one owned audience definition. No delete is exposed over MCP. |
| create_marketing_campaign | Create an owner-private marketing campaign linked only to owned brand/audience ids. |
| update_marketing_campaign | Update one owned marketing campaign. Campaign status never publishes content. |
| create_content_asset | Create a private draft content asset. There is no publish/send operation. |
| update_content_asset | Edit an owned draft. Submitted/approved/rejected revisions are immutable and must be revised. |
| create_content_asset_revision | Create a new owned draft revision from a frozen submitted/approved/rejected revision. |
| submit_content_asset | Submit an owned draft for review, freezing that exact revision and content hash. |
| approve_content_asset | Approve a submitted owned revision and attest to its exact content hash. Requires marketing:approve, separate from draft authority. |
| reject_content_asset | Reject a submitted owned revision with optional reviewer notes. Requires marketing:approve. |
| list_consulting_clients | List consulting clients owned by the API-key account. |
| create_consulting_client | Create an account-private consulting client. |
| update_consulting_client | Update mutable fields on an owned consulting client. |
| list_consulting_engagements | List consulting engagements owned by the API-key account. |
| create_consulting_engagement | Create an engagement for an owned consulting client. |
| update_consulting_engagement | Update mutable fields on an owned consulting engagement. |
| get_consulting_engagement | Get one owned engagement with its document revisions, milestones, and internal deliverables. Long document bodies are previewed. |
| create_consulting_document_revision | Create a revision-safe proposal or SOW draft on an owned engagement. |
| update_consulting_document_revision | Edit an unapproved proposal/SOW revision; approved revisions are immutable. |
| initialize_consulting_milestones | Idempotently initialize stable engagement milestones. Retries return the same ids and external refs. |
| update_consulting_milestone | Update an owned consulting milestone while preserving its stable ref. |
| create_consulting_deliverable | Create an internal draft deliverable on an owned engagement. |
| update_consulting_deliverable | Edit an internal draft deliverable; review/approval state is separate. |
| submit_consulting_deliverable | Submit an owned draft deliverable for internal review without publishing it. |
| reopen_consulting_deliverable | Return an unapproved review deliverable to editable draft status. |
| approve_consulting_document_revision | Approve and lock one exact proposal/SOW revision. Requires the separate consulting:approve scope and never publishes or shares it. |
| approve_consulting_deliverable | Approve one submitted internal deliverable and record approver evidence. Requires consulting:approve and never publishes or sends it. |
| create_workflow_template | Create a new custom workflow template the caller owns, so a multi-step pipeline can be built in conversation instead of by hand in the browser. Saves the same definition the web builder saves. The slug is derived from the name and is returned — use it with run_workflow. Legacy workflows run in dependency order. Explicitly routed workflows use next_step and bounded quality-gate transitions and cannot mix those fields with depends_on. Requires authentication and the workflows:write scope. |
| update_workflow_template | Replace an existing custom workflow template's definition. This is a full replacement, not a patch: supply name and steps as you want them to end up, because anything omitted is not carried over. Built-in templates are read-only, and only the template's owner may change it. Legacy workflows run in dependency order. Explicitly routed workflows use next_step and bounded quality-gate transitions and cannot mix those fields with depends_on. Requires authentication and the workflows:write scope. |
| delete_workflow_template | Delete a custom workflow template the caller owns. Built-in templates cannot be deleted. This removes the saved definition; runs already started from it are unaffected. Requires authentication and the workflows:write scope. |
| create_monitor | Save a standing watch that fires later, on its own: when a count you name crosses a line you set, do the thing you chose. Lets an agent bank a condition and stop polling for it. The watch survives the conversation that created it. Note that start_workflow spends the account's credits unattended each time it fires, so the cooldown is the only thing bounding what it costs. Requires authentication, a Pro or Enterprise account, and the monitors:write scope. |
| list_monitors | List the standing watches you have saved, newest first, with what each one watches, what it does, and when it last fired. Requires authentication and the monitors:read scope. |
| delete_monitor | Delete one of your standing watches, so it stops firing. Does not touch anything it already started. Requires authentication and the monitors:write scope. |
| list_approvals | List workflow checkpoints waiting on you, soonest deadline first, with which step is paused and when it runs out. Defaults to the pending ones. Requires authentication and the workflows:read scope. |
| decide_approval | Answer a waiting checkpoint by request id: approve, reject, or amend. 'amend' means send the step back to be redone with the change described in notes, so notes are required for it. Spends money: approve continues the run and amend re-runs the step, both billed. Fails with a clear error, rather than silently doing nothing, if the run is no longer waiting. Requires authentication and the workflows:approve scope. |
| audit_export | Export this workspace's own agent-action audit trail as structured events, oldest first, for loading into a SIEM. Each event carries what triggered the action, which identity authorized it, which resource was touched, the outcome, and the error state. Metadata only -- no prompt text, document content, tool arguments, or model output is ever returned. Filter by date range and event action; page forward with the cursor each response returns. Covers a documented subset of platform activity rather than every action: see docs/WORKSPACE_AUDIT_EXPORT.md, and note that the coverage block on every response lists exactly which actions are exportable and which are not. Requires authentication and the audit:read scope. |
| account_export | Export this account's own data across every table the account owns, as the same JSON envelope the REST route returns. The set of tables served is derived from the same ownership reflection that decides what an account deletion removes, so a table the account can destroy is a table it can also read. Credentials never appear: columns held encrypted at rest are withheld unless recorded as the account's own content, and password hashes, reset and verification tokens, and API key hashes are withheld by name. Some tables are deliberately excluded -- see docs/ACCOUNT_DATA_EXPORT.md, and note that the coverage block on every response lists exactly which models are served and which are excluded with their reasons. Artifacts are described by their metadata rows rather than inlined. Bounded rather than streamed: each model reports truncated and the envelope reports complete, so narrow to one model and raise rows_per_model rather than re-reading everything. This is a data export, not an erasure request and not a claim of regulatory compliance. Requires authentication and the account:export scope. |
| project_list | List the account's overarching project workspaces, separate from feature epics. Requires tickets:read. |
| project_create | Create an owner-private project with intent and goals. Persist request_id before calling; retry exact input after an uncertain response. Does not start work or create tickets. |
| project_get | Read project intent, current document IDs/revisions and linked feature epics. Read document bodies with project_document_get and follow ticket_get/ticket_list for execution work. |
| project_update | Revision-safely edit project context. Supply expected_revision from project_get and a new request_id. Stale edits conflict instead of overwriting another agent. |
| project_link_feature | Link an existing owned parentless epic to this project. Exact retry reuses the link; linking an epic already belonging to another project conflicts. Does not reparent, claim or execute the ticket. |
| project_unlink_feature | Remove one exact project-feature placement, retaining the ticket. Read binding_id from project_get first; stale bindings or document references conflict. Does not delete tickets or documents. |
| project_document_create | Save the first immutable project brief, PRD, technical specification or decision revision. Optional ticket_id scopes the document to a linked feature. reviewed is editorial metadata, never execution approval. |
| project_document_get | Read one exact owned project document revision and its bounded revision history. Omit revision for latest; use explicit revision when recording evidence. |
| project_document_revise | Append an immutable document revision using expected_revision and a caller-retained request_id. Exact retry reconciles even after later edits; changed input or stale base conflicts. No provider execution or external publication. |