MentisDB
A durable memory engine for AI agents that stores an append-only, hash-chained log of an agent's reasoning plus a version-controlled skill registry, so agents keep what they learned across model swaps and different coding tools.
MentisDB is a durable semantic memory engine and versioned skill registry for AI agents, meant to survive context resets, model changes and staff turnover on a project. It stores semantically typed thoughts in an append-only, hash-chained log through a swappable storage layer, so the memory is tamper-evident rather than a plain file that can be silently edited.
Its skill registry works like a version control system for an agent's instruction bundles: every upload to an existing skill creates a new, cryptographically signable version, none of which are overwritten, and any historical version can be reconstructed. Connecting Claude Code, OpenAI Codex, GitHub Copilot CLI, Cursor, VS Code or any MCP-capable client to the same daemon lets a whole fleet of agents read from and write to one shared memory, and a project manager agent can decompose work across a team of specialists that all draw on it.