LongMemory
A local-first, self-hosted memory engine for LLM applications and agents, with durable SQLite storage, temporal and provenance tracking, and one TypeScript engine reachable from npm, a CLI, HTTP, MCP and VS Code.
LongMemory is a cognitive memory engine for LLM applications and autonomous agents, built around durable local-first storage in SQLite, immutable content with provenance and temporal truth, and several recall modes, including strict, historical, associative and grounded, across multiple languages. It provides governed project memory alongside skills, chat memory, an LLM wiki and a code graph.
A single TypeScript engine backs an npm library, a command-line interface, an HTTP service, an MCP server, a dashboard and a VS Code extension, so the same memory can be used in-process or as a self-hosted server reachable from a zero-dependency Python client. An optional evidence-grounded answering mode validates cited excerpts against authorized evidence and makes at most one model call, while ordinary recall requires none.