Developer docs

Two ways in. Your deployed APIs answer plain HTTPS calls made with a project key. Your workspace (the graph, your tables, entity models, missions and APIs) answers any AI assistant that speaks the Model Context Protocol.

Start here

Connect an AI assistantClaude Code, Claude Desktop, Cursor, Codex or your own agent, over MCP.Call your APIsAuthentication, the request body, errors, limits and long-running work.Admin SDKDeploy and manage your APIs from Python.

1.Authentication

Every call carries a project key in the X-API-KEY header. Create keys in your project under API management. A key belongs to one project, can be limited to some of its APIs, and has a monthly request limit.

X-API-KEY: sk-your-key-here

Keep keys out of URLs and out of code that runs in a browser: anyone who can see a key can use it, and the calls are billed to your account.

2.Calling an API

Open the API in Enact and copy its endpoint: your project's service address followed by /api/run/<api id>. Send the input as the JSON body of a POST.

curl -X POST "https://<your service>/api/run/<api id>" \
  -H "X-API-KEY: sk-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A product photo of a ceramic mug"}'

The body is the API's input itself. To continue a conversation or to label calls in your analytics, wrap it instead: input holds the input, session_id continues a session, tracking_id labels the call.

{
  "input": {
    "prompt": "A product photo of a ceramic mug"
  },
  "session_id": "my-session-1",
  "tracking_id": "order-1042"
}

A successful answer carries the call's calculated cost in the X-Api-Cost header and the time it took in X-Api-Latency-Ms.

3.Errors

An error answers with a status code and a JSON body with a detail field: a sentence, or an object with error and message.

Status	What it means
`400`	The API refused the input. `detail` says why.
`401`	No key was sent.
`402`	Not enough credit, or your plan's monthly usage is used up. `detail.error` is `payment_required`, `detail.message` says what to do and `detail.billing_url` where.
`403`	The key is not valid, or it may not call this API.
`404`	There is no API with this id on this service.
`429`	Too many calls in a short time (wait the number of seconds in `Retry-After`), or the key's monthly request limit is reached.
`500` to `504`	Something failed on our side or at a model provider. Try again after a pause.

A 402 body

{
  "detail": {
    "error": "payment_required",
    "message": "...",
    "billing_url": "/settings?tab=billing"
  }
}

4.Limits

Rate: each key has a short-term rate limit. Above it a call answers 429 with a Retry-After header in seconds.
Monthly requests: each key has a monthly request limit, set per key under API management. Past it calls answer 429 until the next month.
Request size: a request body can be at most 32 MiB. Put larger files in storage first and pass a link.
Time: call your service's own address for long work. The enqualia.io website closes a request after about 60 seconds; the service itself waits for the API to finish.

5.Long-running work

Work in the workspace that takes minutes (missions, ingests, model training) starts a job and answers with its id at once. Follow it with the same key until status is succeeded or failed. Over MCP the same jobs are the get_job tool.

curl "https://enqualia.io/api/qualia/<project id>/jobs/<job id>" \
  -H "X-API-KEY: sk-your-key-here"