We’re officially open-sourcing Anamnesis. GitHub: Anamnesis is a local-first, MCP-native memory bridge that lets your AI memories move across agents. Your memories should not be trapped inside a single agent. Many users have already accumulated a large amount of context across Claude Code, Codex, mem0, Letta, local files, SQLite databases, and various MCP resources: project background, personal preferences, key decisions, debugging history, workflows, and more. But once you switch to another agent, that context often cannot follow you. You may have been working on a project in Codex for a long time, building up rich project context and technical decisions. But when you open Claude Code, it still knows nothing. The memory already exists. It is just locked inside another system. That is why we built Anamnesis. It does not try to create another memory silo. It does not ask you to abandon the agents or memory frameworks you already use. Your original agents continue to generate memory. Anamnesis reads, summarizes, structures, indexes, and serves those memories to other agents through MCP. It can organize fragmented memories from Claude Code, Codex, mem0, Letta, Markdown, local files, SQLite, JSON, and more into a unified structure. Then, any MCP-compatible agent can retrieve and use that context. Agents can change. Context should not be lost. Tools can change. Your memory should always belong to you. The goal of Anamnesis is simple: Let any agent use the memories you have already built up, instead of forcing you to start from zero every time. Your memories should belong to you.