FAQ
Frequently asked questions
Short, factual answers about Keystone — shared organizational memory for AI tools over MCP.
What is Keystone?
- Keystone is shared organizational memory for AI tools. Compatible agents recall source-backed facts, decisions, and knowledge over MCP instead of rebuilding company context in every session.
How is Keystone different from Mem0 or chat memory products?
- Many memory products ask a generative model to silently resolve contradictions. Keystone keeps facts tied to sources and routes rival claims through typed decisions (including TypeSafe System One / Jev). Real conflicts stay visible in an operator inbox instead of being overwritten.
Which AI tools work with Keystone?
- Any MCP-compatible client can connect. We document Cursor, Claude, ChatGPT, and Codex. Connect once; agents in the workspace read the same record.
Does Keystone eliminate hallucinations?
- No. Keystone does not claim to eliminate hallucinations. It gives agents a shared, cited organizational record and flags contested facts so operators can govern what tools recall.
Is there a free plan?
- Yes. Free is enough to connect a tool, ingest a real source, and prove shared recall. Team is for when every agent in the firm should read the same record. Caps are enforced: a call past a limit is refused, not silently billed.
How does conflict handling work?
- When sources disagree on exclusive claims, adjudication scores the rivalry (duplicate, extends, contradicts, or unrelated). Contradictions can open a conflict inbox. List-like facts can coexist. Confidence below threshold does not silently overwrite.
Where is data hosted and how is access controlled?
- Keystone uses Postgres with row-level tenant isolation, scopes, and API keys or OAuth for agents. See the Security page for auth, audit, and infrastructure details. We do not claim SOC 2 or ISO certification at this time.
Still evaluating? Read what organizational memory is, compare Keystone vs Mem0, or contact sales.