What is anamnesis?

Anamnesis is a tool that collects memory from AI agent tools and makes it searchable in one place. If you use several AI assistants or coding agents, each one tends to store what it learned about you in a different format and location. Anamnesis does not ask you to switch memory systems or start over. Instead it reads from wherever your existing agents already store their data. The tool currently supports more than a dozen sources, including Claude Code sessions, Codex memory files, the mem0 framework, and several others. It works by scanning those locations, pulling in the stored notes and context, converting everything to a common format, and saving it into a local SQLite database on your machine. All data stays on your computer. No accounts are required and nothing is sent to a cloud service. Once the memory is imported, you can search it through a command-line tool or by connecting it to another AI agent via a protocol called MCP. The search supports plain keyword matching, similarity-based lookup using local embeddings, or a combination of both. You can filter results by which source they came from, when they were created, or what type of memory record they represent. The project is written in Rust and ships as two programs: a CLI for direct use and an MCP server that lets other agents query your memory automatically. The adapter precision varies by source. Some sources, like mem0 and Letta, extract data with high accuracy when the database schema matches expectations. Others, like Claude Code and Codex, extract at medium precision because those tools use less structured formats. Version 0.1.0 is described as an early release. The README is candid that not every adapter has complete semantic extraction yet, and it lists specific gaps in a limitations section. The Apache-2.0 license covers the code. A Discord community and an X account exist for people who want to follow development or ask questions.