The API

Three credentials reach the API: an account key (rlka_) that acts as you, minted at Account → API keys; a project key (rlk_) bound to one project, minted from the account key or from Integrations → AI tools; and a published site’s own key, carried in the body by page code running in a stranger’s browser. In-app routes take the signed-in session and no key at all. Every card in the reference states which one it takes, and the sample sends exactly that.

A project is the memory boundary every later call is scoped to. An account key at Editor provisions without a browser: create a project per customer or per repository, then mint the narrower keys your services hold.

One dispatch route carries every action. The first call packs code context for a question and answers with the declarations it rested on; Actions lists every family.

curl -X POST https://runlog-7613480744.us-central1.run.app/v1/api/actions \
  -H "Authorization: Bearer $RUNLOG_API_KEY" \
  -H "X-Project-Id: p7Kd2mQx" \
  -H "Content-Type: application/json" \
  -d '{"action": "code_context_pack", "input": {"query": "where is the sweep scheduled"}}'

• Start: keys and projects: everything here runs once, before any data moves.
• Ingestion: what your data becomes: files, working directories, links, schemas, channels.
• Migration: change without re-ingestion: source changes, schema versions, watching, API versions.
• Execution: agents on the memory: retrieval, sessions, artifacts, outbound, business.