Neuronz.ai

English

Français

English

Français

Agent integration (MCP) ​

Neuronz.ai is reached over the Model Context Protocol (MCP). Every capability — memory, recall, rules, knowledge, the graph and more — is an MCP tool, so any agent that can call MCP tools can use the same memory.

How it works ​

Neuronz.ai runs an MCP server over HTTP. Your harness (the coding agent you use: Claude Code, the Claude desktop app, Oh-My-Pi) connects to it in one of these ways:

  • The Claude Code and Oh-My-Pi plugins. The plugin connects the MCP server and also runs the automatic loop: recall on each prompt, the session-start briefing and capture at the end. See Claude Code and Oh-My-Pi.
  • The Claude desktop app, claude.ai and the Claude mobile app. Add Neuronz.ai as a connector and the chat can call the same tools. See Claude desktop app and Claude mobile app.
  • Any other MCP client. Point it at the Neuronz.ai MCP server and it can call the tools directly. Nothing runs automatically there: the agent calls the tools when it decides to.

Features that depend on a harness's hooks differ between harnesses; the harness parity page lists what works where.

You sign in as yourself. When a harness connects, you sign in to Neuronz.ai in the browser. There is no shared token to paste. For automation with no person present, such as a CI job, use an API key instead.

TIP

The full list of tools — what each one does and when to use it — is in the MCP tools reference.