zenbrain

An AI brain that combines zenrag and zenpreference for knowledge retrieval and preference recall. Includes an MCP server and CLI.

Features

• Unified context — Combines knowledge (RAG) and preferences in one getContext call
• MCP server — Model Context Protocol server for Cursor and other AI tools
• CLI — zenbrain for capture, query, and MCP
• Shell history capture — Summarize and index shell history for context

Installation

bun add zenbrain

Set OPENROUTER_API_KEY in your environment for embeddings.

Usage

import { ZenBrain, createEmbeddingModel } from "zenbrain";

const embeddingModel = createEmbeddingModel(); // uses OPENROUTER_API_KEY
const brain = new ZenBrain({ embeddingModel });

// Add knowledge
await brain.addKnowledge(["Your document content here..."]);

// Store preferences
await brain.storePreference({
  key: "packageManager",
  value: "bun",
  content: "Uses bun for package management",
});

// Get combined context for a query
const context = await brain.getContext("How should I run tests?");

CLI

zenbrain capture    # Capture shell history
zenbrain query      # Query knowledge and preferences
zenbrain mcp        # Start MCP server

MCP

Add zenbrain to your MCP config (e.g. Cursor) to expose brain_retrieve, brain_get_context, brain_store_preference, and related tools.

License

MIT