Intake

This file is a stub. The brain-setup skill rewrites it with the real table ids and column ids. If you are reading placeholders below, setup has not run yet.

Four ways material gets into the brain. All of them are optional except the first, which works on day one with no configuration.

1. Conversation

Paste it to Archivist. A thread, a decision, meeting notes, a link. This is the default path and it needs nothing set up.

2. Inbox drop

Write raw material into /brain/inbox/ through the files API and Archivist files it on the next pass.

PUT https://<your-host>/api/data/files/write
Authorization: Bearer <your project API key>
X-Project-UUID: <your project uuid>

{"path":"/brain/inbox/2026-08-06-pricing-thread.md","content":"..."}

The inbox is swept, not triggered. Nothing fires on a file write, so material sits until the next Curator pass flags it or someone asks Archivist to file the inbox. That is a deliberate trade: knowledge capture is not latency-sensitive. If you need it filed immediately, use path 3.

3. Source row

Insert a row into Sources and Archivist wakes on it.

POST https://<your-host>/api/data/tables/<SOURCES_TABLE_ID>/rows
Authorization: Bearer <your project API key>
X-Project-UUID: <your project uuid>

{"cells":[
  {"columnId":"<name>","value":"Q3 pricing decision"},
  {"columnId":"<kind>","value":"thread"},
  {"columnId":"<location>","value":"/brain/inbox/2026-08-06-pricing.md"}
]}

To use this path, enable the table_row trigger on Sources for Archivist in its Automation settings, subscribing to row insertion only. Archivist updates the row it was woken by, so an update subscription would make it wake itself.

4. Ask from outside

Point a Slack command, a bot, or an internal tool at Librarian and get an answer with citations.

POST https://<your-host>/api/agents/Librarian/invoke
Authorization: Bearer <your project API key>
X-Project-UUID: <your project uuid>

{"input":"What is our refund policy for annual plans?"}

To use this path, enable the http trigger for Librarian in its Automation settings. This is a deliberate UI step rather than something Librarian configures itself: Librarian is the agent everyone in the company talks to, and it is not given the ability to change its own configuration.

Repos

If documentation lives in a Git repository, enable the github_repo trigger on push for Archivist. A docs commit then re-files the changed pages.

What does not come in

Attachments. A PDF, a screenshot, or a recording referenced by a source row is a path — nothing reads it. Summarize it in a page by hand, or paste the text.