MemoriLabs/Memori — reverse-engineered prompt

Build me a developer tool called Memori that gives AI agents real memory between sessions. I want it to watch conversations and agent actions, turn them into structured memories, save them, and automatically recall the right details later so users don’t have to repeat themselves.

Make it work with common LLM clients without forcing people to rewrite their agent code. Include a simple Python SDK and TypeScript SDK where someone can register their existing client, set a user id and process id, then keep making normal chat calls while memories are saved and recalled in the background.

Also include a cloud API key setup, environment variable examples, local development setup, tests, and clear docs with quickstarts. Add examples for remembering facts like a favorite color, plus integrations for agent tools like MCP, OpenClaw, and Hermes if practical. It should be datastore flexible, with an option for people to bring their own database later. Look up current docs online if you need to.