Facts and Wiki

The canonical memory surface has two resources:

/admin/facts reads the fact ledger and accepts explicit user pins.
/admin/wiki/pages manages versioned wiki pages.

All endpoints require the usual Bearer token. See Authentication.

Facts

List and search

GET /admin/facts lists facts. GET /admin/facts/search?query=... searches them. Both accept subject, include_inactive, and limit (1–100). Use the returned next_after object’s created_at and fact_id as after_created_at and after_fact_id together to continue a page.

curl "http://localhost:6877/admin/facts/search?query=coffee&limit=20" \
  -H "Authorization: Bearer $API_KEY"

Responses contain facts, total, has_more, and next_after. Read one fact with GET /admin/facts/{fact_id}; its response contains fact.

POST /admin/facts persists one explicit user pin. Send a stable request_id to make retries idempotent:

curl -X POST http://localhost:6877/admin/facts \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"request_id":"result-2026-07-30-1","text":"Keep this result."}'

Other fact changes use the plan-and-commit agent tools rather than generic HTTP mutation endpoints.

Batch read

POST /admin/facts/batch reads up to the server’s batch limit in one request:

curl -X POST http://localhost:6877/admin/facts/batch \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"fact_ids":["fact_1","fact_2"]}'

Wiki pages

These /admin endpoints are the Desktop/backend concurrency interface, so they expose internal page and revision identifiers. Agent wiki tools do not: they use managed paths and keep this bookkeeping inside the server.

Read

GET /admin/wiki/pages lists active pages. Pass include_redirects=true to include redirects.
GET /admin/wiki/pages/{page_id} reads one page, including content.
GET /admin/wiki/pages/{page_id}/history?limit=50 returns commit history (1–200).
GET /admin/wiki/pages/{page_id}/diff returns a diff. It accepts optional base, target, offset, limit, and tail.
GET /admin/wiki/pages/{page_id}/links returns outgoing links and backlinks.

Page responses include page_id, path, resource_state, title, excerpt, aliases, lifecycle, redirect_to, metadata, version, repository_head, created_at, and updated_at. Detail responses also include content.

Create

POST /admin/wiki/pages requires path, title, and expected_head. It also accepts body, page_id, aliases, and metadata.

curl -X POST http://localhost:6877/admin/wiki/pages \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"path":"notes/coffee.md","title":"Coffee","body":"# Coffee\n","expected_head":null}'

Update, archive, and restore

PUT /admin/wiki/pages/{page_id} requires content, expected_version, and expected_head. Archive and restore use POST and require expected_version and expected_head:

curl -X POST http://localhost:6877/admin/wiki/pages/page_123/archive \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"expected_version":"rev_123","expected_head":"commit_123"}'

Writes are optimistic: supply the version and repository head you read. A stale write returns 409 with the current revision, head, and, when available, content. Invalid paths or request values return 422; unknown resources return 404. Restore one historical revision with:

POST /admin/wiki/pages/{page_id}/history/{commit_id}/restore

The request uses the current expected_version and expected_head, just like other wiki writes.

Overview

Endpoints

Facts and Wiki

Facts

List and search

Pin

Batch read

Wiki pages

Read

Create

Update, archive, and restore

​Facts

​List and search

​Pin

​Batch read

​Wiki pages

​Read

​Create

​Update, archive, and restore

Facts

List and search

Pin

Batch read

Wiki pages

Read

Create

Update, archive, and restore