Extended context memory that works seamlessly with your agent's own context, expanding its storage beyond the context window it comes with. Every agent you run shares it: Claude, GPT and Grok agents read what the others saved. neomonic free is one small MCP server for Claude Code, Cursor or any MCP client: a single file, no installer, nothing to sign up for, and everything kept word for word on your own machine.
The free edition gives you one context, named neomonic. Topics inside it keep things apart.
No Ollama, no GPU, no Python, no cloud account and no API key: the one file is the whole memory. All three hashes are also in SHA256SUMS.
Unpack it anywhere, then register it with your agent.
claude mcp add neomonic -- /path/to/0neo mcp
On Windows, give the full path to 0neo.exe. Restart Claude Code and ask it to remember something.
{ "mcpServers": { "neomonic": { "command": "/path/to/0neo", "args": ["mcp"] } } }
Run one server and point every agent at it. Each agent embeds the memory in its own keyphrases, and all of them read and write the same store.
0neo mcp --http 9966 claude mcp add --transport http neomonic http://127.0.0.1:9966/mcp
Tested with Claude, GPT and Grok agents writing at the same time: each answered every question about facts only the other two had saved.
xattr -dr com.apple.quarantine neomonic-free-macos-universal
The binary is signed ad hoc, not notarized, so macOS says the developer cannot be verified until the quarantine tag is removed.
Each agent embeds the memory in its own keyphrases: it registers under its model, files every memo under its own headings and cues, and recalls through them. Claude, GPT and Grok each build their own way in over the same memos, and none of it needs a GPU or an outside service. Memory is kept in ~/.neomonic/contexts (Windows: C:\ProgramData\neomonic\contexts).
Start free. Move up when one context is not enough.
neomonic