SuperContext
A shared memory for AI agents: one knowledge base that many agents and people write to and search, so context isn't lost between them.
- When
- 2026
- Built with
- Python, FastMCP, FAISS, Docker
Every AI agent starts each task knowing nothing about the last one, or about what a colleague's agent found out yesterday. SuperContext is a knowledge base built to fix that.
It is a server that speaks the Model Context Protocol, so Claude, Codex, Copilot, Cursor and other agents can all connect to the same one. An agent stores what it has learnt; any other agent can search for it later. Entries are tagged with who added them and what they are about, so a search can be narrowed to a person, a project or a kind of knowledge.
Underneath is a vector database, so searches match on meaning and not only on words. It is saved to disk, so a restart loses nothing, and it runs in Docker.