Each successful call decrements a monthly call allowance by one, regardless of token count. Plans set the monthly cap: dev 200, startup 20000, scale 200000. Token counts are recorded per call for cost analysis.

Meeting Intelligence API

Q: How do I authenticate?

Send your API key on every v1 request as an Authorization Bearer header or an X-API-Key header. Keys are shown once at creation and stored only as a hash.

Q: What happens when I hit my monthly cap?

The API returns HTTP 429 with error quota_exceeded. The counter resets at the start of each calendar month. Upgrading the plan raises the cap and takes effect on the next call.

Send meeting transcript text and get back a running summary, a consolidated overview, or an answer grounded in the conversation. Transcription stays in the browser; this API is the server-side AI over the text. Authenticated by API key, metered per organisation.

Get an API key

On this page: Authentication · Endpoints · Generate · Usage · Keys · Plans · Errors · FAQ

Base URL is your Worker domain. Every request and response is application/json.

Authentication

Send the API key on every /v1 request, in either header form:

Authorization: Bearer lt_live_XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
X-API-Key: lt_live_XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Mint keys at the key dashboard or with POST /api/keys. Only the SHA-256 hash is stored; the raw key is shown once. lt_live_ and lt_test_ keys both bill against the same organisation counter and differ only by label. Key management is authenticated by your account access token, not by an API key.

Endpoints

Method	Path	Auth	Purpose
POST	`/v1/generate`	API key	Summary, overview, or Q&A over a transcript
GET	`/v1/usage`	API key	Plan, calls used, remaining, tokens used
POST	`/api/keys`	Account token	Mint a key (returned once)
GET	`/api/keys`	Account token	List keys and organisation summary
DELETE	`/api/keys/:id`	Account token	Revoke a key

POST /v1/generate

Request body fields:

Field	Type	Required	Notes
`type`	string	yes	`summary`, `overview`, or `qa`
`text`	string	summary, overview	Transcript text. Truncated at 60,000 characters
`question`	string	qa	The question or drafting request. Truncated at 2,000 characters
`history`	array	no	`[{ "role": "user"\|"model", "content": "…" }]`. Last 20 turns kept
`outLang`	string	no	English, Japanese, Portuguese, Spanish, German, French, Italian

`type` behaviour

type	Output
`summary`	One to two short lines summarising a segment, roughly 25 words
`overview`	Consolidated bullet overview of the full transcript, with a decisions and actions block when present
`qa`	Answers a question grounded in the transcript, or drafts the actual words to say when asked to

Request

curl -sS https://<worker>/v1/generate \
  -H "Authorization: Bearer $LT_KEY" \
  -H "content-type: application/json" \
  -d '{
    "type": "qa",
    "text": "…transcript so far…",
    "question": "Draft a one-line closing remark that confirms the next step."
  }'

Response `200`

{
  "text": "…model output…",
  "usage": { "promptTokenCount": 812, "candidatesTokenCount": 96, "totalTokenCount": 908 },
  "meta": { "plan": "startup", "cap": 20000, "remaining": 19860 }
}

Response headers report the meter: X-RateLimit-Limit, X-RateLimit-Remaining, and X-Usage-Total-Tokens. Each successful call decrements the monthly allowance by one, regardless of token count.

GET /v1/usage

curl -sS https://<worker>/v1/usage -H "Authorization: Bearer $LT_KEY"

{
  "plan": "startup",
  "calls_used": 140,
  "cap": 20000,
  "remaining": 19860,
  "tokens_used": 1284310,
  "period_reset_at": "2026-09-01T00:00:00+00:00"
}

Key management

Obtain the account token with a Supabase password grant, then call the Worker. Most teams simply use the dashboard instead.

# 1. sign in
TOKEN=$(curl -sS "$SUPABASE_URL/auth/v1/token?grant_type=password" \
  -H "apikey: $SUPABASE_ANON" -H "content-type: application/json" \
  -d '{"email":"[email protected]","password":"…"}' | jq -r .access_token)

# 2. mint a key (returned once)
curl -sS https://<worker>/api/keys \
  -H "authorization: Bearer $TOKEN" -H "content-type: application/json" \
  -d '{"name":"production-backend","mode":"live"}'

# 3. list keys
curl -sS https://<worker>/api/keys -H "authorization: Bearer $TOKEN"

# 4. revoke a key
curl -sS -X DELETE https://<worker>/api/keys/<id> -H "authorization: Bearer $TOKEN"

Plans and limits

Plan	Calls / month	Soft rate (req/min)
`dev`	200	20
`startup`	20,000	120
`scale`	200,000	600

The monthly cap is enforced atomically in the database and resets at the start of each calendar month. The per-minute rate is a best-effort guard, not a billing boundary.

Errors

Every error is { "error": "<code>" } with the HTTP status below.

Status	error	Cause
400	`invalid_json`	Body was not valid JSON
400	`unknown_type`	`type` not in summary, overview, qa
401	`api_key_required`	No key on a `/v1` request
401	`invalid_key`	Key not found or revoked
403	`org_suspended`	Organisation status is not active
429	`quota_exceeded`	Monthly call cap reached
429	`rate_limited`	Per-minute soft guard tripped
502	`ai_unavailable`	Upstream model error

FAQ

Do I send audio to the API?

No. The API operates on transcript text you supply. No audio reaches the server, and transcript text is not stored; only token counts and call metadata are recorded for billing.

How is usage billed?

Each successful call decrements a monthly call allowance by one, regardless of token count. Token counts are recorded per call so you can analyse cost. See plans for the monthly caps.

How do I authenticate?

Send your API key on every /v1 request as an Authorization: Bearer header or an X-API-Key header. Keys are shown once at creation and stored only as a hash.

What happens when I hit my monthly cap?

The API returns 429 with quota_exceeded. The counter resets at the start of each calendar month, and upgrading the plan raises the cap on the next call.

How do I rotate a key?

Create a new key, deploy it, then revoke the old one. Revocation takes effect immediately for the next call.