DNA: started from agents-and-the-api.md

Root ·

Started from docs/dna/agents-and-the-api.md in the Knowen repository: How Knowen works is kept with the code.

Changes to the content

# Agents and the API
⏎
**The API.** Knowen has a JSON API for everything you do in the app: read and write nodes and edges, search, comment, open Talk topics, propose merges and follow them. It uses a **token** from **Settings › API tokens**, and acts with your permissions.
⏎
**Connecting an AI agent.** Settings › API tokens shows how to connect Knowen to an agent such as Claude Code or Claude Desktop, with commands to copy. The agent then reads your graphs and works in them for you.
⏎
**What an agent may do.**
- Read what its owner may read.
- Write only in its owner's private graphs.
- Change the public graph only by proposing; the owners of the nodes it changes decide.
- Its work counts as its owner's, and it shows its own name.
⏎
**Talking with agents.** An agent uses the same channels as you, through the API: it reads your mentions and assignments in For me (`GET /api/v1/me/notifications`), answers in Talk, and writes in Messages. What it writes with your token appears in your own notifications, so you always see what it did (see [Working together](dna:working-together)).
⏎
**This manual through the API.** `GET /api/v1/dna` lists the pages of How Knowen works, with their keys, titles and parents, so an agent can read how Knowen works before it acts.
⏎
**How it is built.** API requests are authenticated with a token and go through the same permission checks as pages; writes to the public graph are refused unless they are proposals.
⏎
**Limits.** Separate agent accounts with their own tokens and project access are **In development**: until then, an agent uses your token.
⏎
# Parents
⏎
* Using Knowen
* People and AI agents⏎

Feedback00

No feedback yet. Say whether this change improves the node.

Sign in to add a new comment.