Acta

Appearance

For developers ​

This half of the documentation is for running Acta, integrating with it, or working on it. If you are using an Acta someone else runs, you want the users' half instead.

Running it ​

Acta is one process: API, MCP endpoint and web app on a single port, with SQLite and a directory of attachments. No separate database server, queue or cache.

  • Install: Docker, behind a reverse proxy, on Cloudflare Workers, or from source.
  • Configuration: every environment variable, read out of the code rather than from memory.
  • Authentication: email one-time codes, OpenID Connect against your own provider, and access tokens.

Moving in ​

  • From Trello: boards, cards, comments, attachments and members, with a dry run and a reconciliation first.
  • From Confluence: page trees, version history and cross-references, including an honest list of what does not convert.

Building on it ​

  • The MCP endpoint: what it is, why it is shaped for token economy, and the two kinds of token.
  • Tools: the tool reference.

The REST API under /api/v1 mirrors the same services, so anything an agent can do a script can do.

Working on Acta ​

  • Architecture: the two runtimes, the write path, and the three things that catch people out.
  • The data model: the tables that carry the ideas.

Contributions go through the contributing guide and an Individual CLA.

Before your first schema change

A failed migration is memoized per isolate, so a bad one is not a degraded feature: it is every request failing until a redeploy. Read the migrations section of Architecture first.