BaseMouse

About BaseMouse

BaseMouse gives AI agents one shared, versioned memory across Claude Code, Cursor, Codex, and Grok through a single remote MCP server. The problem it names is familiar to anyone using more than one coding agent: project context ends up trapped per tool — a CLAUDE.md here, a Cursor rules file there — and the whole project gets re-explained to every new agent. BaseMouse replaces that with one corpus every registered tool reads from and writes to, so whatever an agent learns in one tool is available to an agent in another, even a different vendor; only the live chat stays inside each tool. Setup is a folder of documents pointed at the service, which reads them, keeps every version, and labels them for retrieval, with re-imports that do not duplicate and a clear report on any file that fails to import. Retrieval returns the most relevant passages with source, a tamper-evident fingerprint, and recency, ready to drop into a prompt — and a get_context_pack tool returns an agent-shaped bundle with citations. Governance is a first-class concern: every change is recorded, nothing is silently overwritten, and even deletions stay on the record, so you can always answer where an answer came from and what changed since. The model is open core — the engine that imports documents and serves sourced answers is free and MIT-licensed and self-hostable with nothing phoning home, while the paid hosted version handles running, billing, and support.

Key Features

BaseMouse Pros & Cons

Who Is BaseMouse Best For?

Tags

More AI Agent Infrastructure tools