Semantic search with citations
Vector retrieval and token budgets shape concise context. Every result includes its source, making answers verifiable.
SEARCH / QUERY / ASKEnterprise knowledge infrastructure built for AI agents
Give every agent
one shared company brain
Turn scattered meetings, customer context, and decisions into searchable, citable knowledge that evolves with your company. Agents search first and learn while they work.
markdown-in-git is the source of truth · every index is rebuildable
Company knowledge graph128 pages · synced
ACME overviewwiki / customer
07-14 customer callraw / evidence
Q3 renewal planwiki / plan
Key contactswiki / people
Product feedbackraw / interview
Built on open, portable foundations
Markdown
Git
PostgreSQL
pgvector
MCP
OAuth 2.1
How it works
Knowledge grows from real work
No separate knowledge chore. Agents retrieve evidence, preserve conclusions, and leave every change traceable while completing actual work.
01 · Search before answering
Retrieve real company knowledge with a source for every claim. No evidence means no guess.
02 · Separate evidence and insight
raw preserves immutable source material; wiki holds knowledge and judgments agents can maintain.
03 · Evolve knowledge safely
Natural-language edits, atomic multi-page changes, and a complete Git history for every write.
AGENT SESSION● CONNECTED
# The agent identifies company context first $ cbrain me → company: acme · role: editor $ cbrain query "Q3 renewal risks" → 6 relevant results · every result cited
Capabilities
Designed for agent collaboration
Lightweight, open, and secure by default. Keep ownership of your knowledge while giving every agent reliable context.
Open and portable
Markdown files and Git history are the truth. Databases, vectors, and indexes are derived caches you can rebuild at any time.
OKF / MARKDOWN / GITSelf-maintaining knowledge
Agents preserve meetings, customer changes, and decisions as they work. Incremental indexing and digests keep knowledge fresh.
INGEST / DIGEST / REBUILDConnect any agent
Use the CLI, HTTP, or MCP with Cursor, Claude, Codex, or internal agents. Authorize once; permission changes apply instantly.
CLI / HTTP / MCPSecure by structure
If you cannot see it, we do not reveal it exists
cbrain protects company knowledge with structural boundaries. Every read must pass tenant, clearance, and department checks. Access is denied by default.
✓Tenant isolation through application guards and PostgreSQL RLS
✓Team, company, and administrator clearance levels
✓Department-scoped directory boundaries for reads and writes
✓OAuth 2.1, PKCE, short-lived tokens, and complete audit trails
Your company knowledge, agent-ready
Give your team one shared brain
Start with one trusted fact and help every agent understand your company.
Open the cbrain workspace