# Open Lab

> A LessWrong-style discussion forum built on ATProto (the Bluesky protocol).
> Every post, comment, vote, and reaction is a record in the author's own data
> repository (PDS), published with the open `site.standard.*` lexicons. The
> site is an index that aggregates those records.

## When to use Open Lab

- You want to read or discuss research-adjacent writing in a LessWrong-style
  community, without locking content into a private database.
- You are an ATProto user who wants your posts stored in *your* PDS and remain
  portable to other apps (leaflet.pub records appear here too).
- You want a worked example of an ATProto-native web app: OAuth login,
  Jetstream ingestion, lexicon-driven data, karma via `site.standard.graph.recommend`.
- Not for: real-time chat, private discussions, or content you do not want public.

## How to read Open Lab as an agent

- Send `Accept: text/markdown` with any request and every page returns its
  markdown variant (`text/markdown; charset=utf-8`, with `Vary: Accept`).
- Key pages:
  - `/` — frontpage ranked by karma
  - `/allPosts` — everything chronologically
  - `/concepts` — posts grouped by tag (`/concepts/<tag>` for one tag)
  - `/library` — publications (blogs)
  - `/posts/<did>/<rkey>` — a single post
  - `/users/<actor>` — an author profile (handle or DID)
  - `/about`, `/contact`, `/privacy`

## Machine interfaces

- MCP server (Streamable HTTP JSON-RPC): `/.well-known/mcp`
  - tools: `get_frontpage`, `get_post`, `list_concepts`
- OpenAPI description of HTTP endpoints: `/openapi.json` (stable contract at `/api/v1/*`; errors are RFC 9457 problem+json; rate-limit headers on every API response)
- API catalog: `/.well-known/api-catalog.json`
- Sitemap: `/sitemap.xml`

## Policies

- Sign-in uses ATProto OAuth; we never see app passwords. One session cookie
  holds your DID. No analytics, no trackers.
- To remove content from this index, delete it at the source PDS, or email
  research@protocol.ai.