FAQ

What is AgenticMind?

An auditable, self-improving knowledge & memory layer for AI agents, served over MCP. Answers are citation-enforced (no source, no claim) and carry a replayable why-trace. It's self-hosted on Postgres alone, Apache-2.0, and maintained by Moai Team LLC.

How is it different from a vector store or memory SDK?

Plain memory gives you fuzzy recall and zero accountability. AgenticMind refuses unsupported answers, keys every claim to a numbered source, records a replayable why-trace, and promotes validated knowledge through a judge-gated compounding loop. The differentiator is provenance and governance, not raw recall — see Provenance is the differentiator.

How do I run it?

One command, no clone:

OPENAI_API_KEY=sk-... sh -c "$(curl -fsSL https://raw.githubusercontent.com/Moai-Team-LLC/AgenticMind/main/quickstart.sh)"

It needs Docker and an OpenAI-compatible key. Embeddings run locally with zero keys.

Does my data leave my infrastructure?

No. AgenticMind is self-hosted and runs embeddings locally (bge-m3), so ingested text stays in your Postgres. Only the optional synthesis step calls your configured chat model.

Is it secure and multi-tenant?

The MCP endpoint is fail-closed and bearer-gated with least-privilege, revocable scopes, and every request runs inside a Postgres row-level-security tenant context. See the security model.

What is the license and who maintains it?

Apache-2.0, maintained by Moai Team LLC. AgenticMind is the reference implementation of the Agentic Product Standard.