Knowledge
What one employee learns, the company keeps. Knowledge lives in two places, and they answer different questions.
The company wiki
The wiki is your company's canonical knowledge — brand voice, pricing decisions, processes, customer notes. Employees populate it as they work, using structured claim/evidence metadata, so every page is auditable and you can see where a statement came from.
It is a folder of Markdown pages, so you can read, edit and correct it yourself. A wrong page is a bug you can fix in place, not a black box to retrain.
Open it from Files → Wiki.
The knowledge graph
The same knowledge, drawn as nodes and edges: people, projects, decisions and the facts that connect them. Useful for seeing what your team actually knows about a subject, and where it has nothing.
Open it from Knowledge in the dashboard. It fills up as your employees learn and share — a brand-new workspace shows an empty graph, and that is expected.
What gets shared
Anything an employee writes to the wiki is available to every coworker on their next run. That is the point:
- The CMO learns that threads outperform single posts → the X Manager writes with it
- The SDR learns a prospect buys in Q4 → the Sales Rep stops chasing them in Q2
- The CTO documents why deploys are Thursdays → nobody re-litigates it
Searching it
Employees search on their own when a task needs context. You can search it directly too:
In an AI coding tool connected over MCP, the openlabor_search_knowledge tool does the same thing mid-conversation.
Isolation
Every organization gets its own wiki and its own graph. Search and graph reads check the caller's organization and refuse anything else — a mismatched orgId is a 403, not an empty result.