Brain Cartridge Memory System
Self-contained semantic search without the LLM. Click a cart to mount it, then search with plain English — no generation, no hallucinations.
Search your brain cartridge
Mempack — portable per-agent memory. Each agent gets its own writable cart that travels with it. This dashboard is your home base for what your agent has been doing on your behalf.
Your dispatched tasks sit in this Mempack until an MCP-aware client mounts it. Drop the snippet below into your client's mcp.json, restart it, then ask the agent to mount your Mempack and check for dispatches.
%APPDATA%\Claude\claude_desktop_config.json (Win) / ~/Library/Application Support/Claude/claude_desktop_config.json (Mac)~/.cursor/mcp.json.mcp.json (project) or ~/.claude/mcp.json (global)~/.codeium/windsurf/mcp_config.jsonMempack Local Agent is a small Python script that wires a local Ollama model to your Mempack via MCP. No Claude subscription, no API key — just Python + Ollama. Bring your own model; the script handles the tool-call loop.
ollama pull qwen2.5:14b # or hermes3:8b, qwen2.5-coder:14b, llama3.1:8b pip install requests truststore # truststore only needed on Win w/ corporate AV
Download Ollama: ollama.com/download
requests) + optional truststore