API

There are two working paths, and they authenticate differently.

Survey submissions — POST https://api.annsa.ai/surface/submit. Take the key from Settings → Integrations → Survey and send it in the JSON body as api_key, not as a header.

{
  "api_key": "YOUR_SURVEY_KEY",
  "text": "The export keeps timing out on large boards",
  "email": "sam@acme.com"
}

Everything else — use MCP. The remote server at https://app.annsa.ai/mcp authenticates over OAuth in the browser and exposes priorities, specs, actions and search to any MCP client or agent. See Cursor & Claude Code.

There is no general-purpose REST API key today. The in-app endpoints authenticate with your session, so they are not usable from a script or an agent.

Same endpoint, an items array instead. Up to 1,000 items per request, and the response carries a result for each one.

{
  "items": [
    {
      "text": "Can't find the export button",
      "customer_email": "alex@example.com",
      "revenue_band": 49
    },
    {
      "text": "Love the new dashboard layout",
      "customer_name": "Sam Lee"
    }
  ]
}

What /surface/submit accepts:

Richer customer fields — name, revenue, SKU, your own external id — are set on import (CSV, Google Sheets) or over MCP. See Importing feedback.

Annsa hashes the text, so the same content submitted twice creates one item. Pass external_id and the second submission is skipped on that id regardless of what the text says.

A new item returns its id:

{ "id": "8ffee9c7-63e6-4e5d-9fb3-387a97bbdc0c" }

A duplicate says so, and names where the original came from:

{
  "id": "8ffee9c7-63e6-4e5d-9fb3-387a97bbdc0c",
  "duplicate": true,
  "original_source": "slack"
}

A batch returns counts and a result per item — total_items, successful, queued, duplicates, failed, plus results, duplicates_list and errors.

5,000 requests per hour on survey submissions, and 1,000 items per batch request. Your plan's monthly feedback limit applies separately — past it, items are safely held rather than dropped. See Held feedback & plan limits.

Feedback submitted this way is tagged widget, so you can filter it in the activity log and tell it apart from Slack, CSV and MCP submissions. The source is set for you — you don't pass it.

Every endpoint here is described in an OpenAPI 3.1 document at annsa.ai/openapi.json — including which credential each one accepts, so an agent can work that out without reading this page. It is served with an open CORS header, so a tool can fetch it directly.

It is also linked from every response on the site as Link: </openapi.json>; rel="service-desc".

POST https://api.annsa.ai/transcripts/upload

Uploading the same transcript twice is safe; it is recognized rather than duplicated.