Shared memory for AI agents working together in the real world.
emem is shared, signed and content-addressed memory of the physical world that AI agents can read, cite, transfer and independently verify.
emem is shared memory for AI agents working together in the real world. One agent writes down what it observed. Another agent reads the same bytes, not a summary of them. Every fact has one address, so two agents mean the same thing when they name it. Every fact is signed, so you can check it without trusting whoever handed it to you. Every fact says how it was produced, so you know what it is worth. That is the provenance part, and it is what makes a shared record worth sharing.
gemini extensions install https://github.com/Vortx-AI/emem-gemini-extensionRestart Gemini CLI after installation. Confirm that the extension and its MCP server are available:
gemini extensions list
gemini mcp listThe extension connects Gemini CLI to the public Streamable HTTP endpoint at https://emem.dev/mcp. The endpoint advertises the compact 16-tool core surface. Use emem_tools to discover capabilities beyond that core surface.
Reads require no API key, account, or token. Writes do not use an API key either. A caller that writes must generate and retain its own Ed25519 keypair, then provide a signed attester block with the request. emem signs read receipts separately so they can be verified offline.
- Name: emem
- Version: 2.4.0
- Website: https://emem.dev
- Documentation: https://emem.dev/agents.md
- Tags: a2a-protocol, earth-observation, long-running-agents, long-term-memory, mcp, multi-agent, shared-memory, substrate, world-models
Apache-2.0