Content agent - CaptureFlow API

The content agent is the same agent that writes drafts in the app. Through the API it runs in the background: you start a run, then read it until it’s done. Nothing is posted; the drafts wait for you in CaptureFlow. Scope: agent:run.

Start a run

POST /v1/agent/runs with exactly one of:

Field
`idea_id`	One of your ideas.
`input_id`	One of your inputs (a transcript, an article, a document).
`text`	20 to 100,000 characters of notes, a transcript or an article. Saved as an input; `title` names it.

Optional instructions (up to 2,000 characters) steer the drafts: tone, length, the angle to take.

curl https://api.captureflow.ai/v1/agent/runs?workspace_id=$WORKSPACE_ID \
  -H "Authorization: Bearer $CAPTUREFLOW_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{ "input_id": "'$INPUT_ID'", "instructions": "Two short LinkedIn posts, practical tone." }'

If that input already has a background run, it is returned with reused: true and the new instructions are not used. A reused run still counts toward the daily limit.

Follow it

GET /v1/agent/runs/{id} until status is completed (or failed); it usually takes a few minutes. The finished run lists its drafts (up to 50) and app_url, where you review them and turn one into a post.

Limits

One run at a time per person in a workspace: agent_run_in_progress, with details.run_id.
20 runs per person per workspace per day: agent_daily_limit.
A daily limit for all background runs of a workspace: agent_workspace_limit.
The agent may be switched off for a workspace: agent_disabled.
In an agency client workspace, members can’t start runs (managed_member_restricted).
Agent calls have their own, tighter rate limit.