CrowTerminal

Delete a transcript via DELETE /api/desktop-sync/transcripts/{id}. DESTRUCTIVE. A soft delete at CrowTerminal -- the record stops appearing in the listing and stops counting, and this API offers nothing that restores it. Measured 2026-09-17: the first call answers 200 {"success":true} and the second answers 404 "Transcript not found", so a repeat is not idempotent.

Get the current user via GET /api/auth/me. The account the connection acts as: id, email, name, avatar, plan (FREE/STARTER/GROWTH/AGENCY/ENTERPRISE), auth provider and counts of related records.

Compare a client against its niche via GET /api/benchmarks/client/{clientId}. Percentiles and vs-average figures for one client against the benchmark for its niche.

Get a client's benchmark history via GET /api/benchmarks/client/{clientId}/history. Percentile snapshots over time, newest first; `limit` defaults to 12.

Get benchmarks for a niche via GET /api/benchmarks/niche/{niche}. `niche` is matched loosely -- "Fashion & Lifestyle" resolves to the `fashion` benchmark. A niche with too little data answers 404 with "Benchmarks require at least 5 accounts and 20 videos", which is a data condition rather than a bad path.

List niches with benchmark data via GET /api/benchmarks/niches. Each entry carries the niche key to pass to the per-niche tool, a display name, how many accounts back it and a confidence figure.

Get a benchmark summary across all clients via GET /api/benchmarks/summary. Per-client percentiles plus agency-wide totals (client count, average percentile, top performers, clients needing attention).

List clients via GET /api/clients. Paginated. `search` matches the client name and `isActive` filters on the active flag; both were measured working 2026-09-17. Unknown query parameters are ignored rather than refused.

Get a client via GET /api/clients/{id}. `id` must be a UUID; any other shape answers 400 "Invalid client ID" from the request validator rather than 404.

Count synced transcripts via GET /api/desktop-sync/count. The number of transcripts held in the cloud for this account, excluding soft-deleted ones.

List desktop sessions via GET /api/desktop-sync/sessions. Registered desktop installations for this account: device id, platform, app version and when each was last seen.

List synced transcripts via GET /api/desktop-sync/transcripts. Transcripts uploaded by the CrowTerminal desktop app, newest first; `limit` defaults to 50 and `clientId` narrows to one client. Soft-deleted transcripts are excluded.

Get prediction accuracy summary via GET /api/predictions/accuracy/summary. Aggregate error between predicted and recorded actuals across every client. Answers 200 with a null summary and an explanatory message while no actuals have been recorded.

Get a client's prediction history via GET /api/predictions/{clientId}/history. Stored predictions for one client, newest first, each with its id, predicted figures and any recorded actuals. This is where the id for the single-prediction and record-actuals tools comes from.

Get a prediction via GET /api/predictions/{id}. `id` is a prediction id, not a client id. An unknown one answers 404 RES_001 naming the resource; a non-UUID answers 400 VAL_001 "Invalid prediction ID".

Get cost projection via GET /api/usage/cost-projection. This account's month-to-date cost and the projection for the full month, with the day counts the projection is built from.

Get usage quota via GET /api/usage/quota. Plan, monthly token limit, tokens used and remaining, reset date and whether overage is allowed. This is the budget the engagement-prediction tool spends.

Get usage summary via GET /api/usage/summary. This account's own usage over `days` (1-365, default 30): tokens, requests, success rate, latency, estimated cost, and breakdowns by feature and by day. `days` outside the range answers 400 "Days must be 1-365".

Update the current user's profile via PATCH /api/auth/me. Accepts `name` and `avatarUrl` only. It cannot change the email or the password, so it cannot lock this connection out of the account.

Record actual results for a prediction via PATCH /api/predictions/{id}/actual. `actualEngagement` (0-100) and `actualViews` are required. Returns the error between predicted and actual and feeds the accuracy summary.

Log in to an existing account via POST /api/auth/login. Exchanges any CrowTerminal email and password for a 7-day JWT and returns that token in the reply. This connection does NOT need it: Agentic Fabriq performs the same exchange itself from the login stored on the connection and attaches the result to every other tool. Rate limited by CrowTerminal to 5 attempts per hour per source IP (measured 2026-09-17: ratelimit-policy 5;w=3600), a budget shared with the connection's own automatic sign-in, so calling this tool can leave the connection unable to renew its session for up to an hour.

Register a new user via POST /api/auth/register. Creates a NEW CrowTerminal account from an email, password and name -- it does not act on the account this connection holds, and the account it creates is not connected to Agentic Fabriq. The reply carries a 7-day JWT for the new account. Rate limited by CrowTerminal to 3 registrations per hour per source IP (measured 2026-09-17: ratelimit-policy 3;w=3600), which is shared by every connection on this deployment.

Create a Stripe checkout session via POST /api/billing/checkout. PRIVILEGED. Starts a paid subscription purchase for the connected account and returns a live Stripe Checkout URL; anyone holding that URL can complete the purchase. `priceId` is a Stripe price id -- an unknown one answers 400 "Invalid price ID" (measured 2026-09-17).

Create a Stripe billing portal session via POST /api/billing/portal. PRIVILEGED. Returns a live Stripe customer-portal URL for the connected account, carrying a session secret in the query string. Anyone holding that URL can change or cancel the subscription and see the stored payment methods. Measured 2026-09-17: answers 200 with such a URL on a FREE-plan account.

Create a client via POST /api/clients. `name` is the only required field. The enumerated fields (`brandVoice`, `contentGoals`, `language`, `postingFrequency`, `region`, `targetGender`) accept only the listed values. Returns the created client including the `id` the benchmark and prediction tools take.

Register a desktop session via POST /api/desktop-sync/sessions. `deviceId`, `platform` (darwin/win32/linux) and `appVersion` are required. Re-registering a known device updates its last-seen time rather than creating a second row.

Upload a transcript via POST /api/desktop-sync/transcripts. `fileName`, `text` and `language` are required; `clientId`, `duration`, `localId` and `createdAt` are optional. Returns the stored transcript's id.

Predict engagement for planned content via POST /api/predictions/engagement. BILLED IN TOKENS against the account's monthly quota (see the usage-quota tool). Takes a `clientId` and a `content` object (caption, hashtags, posting time, video duration) and returns a predicted engagement rate with a range, a view range, an overall score and per-signal breakdowns. The prediction is stored and gets an id, which the prediction-history tool reads back.