Who it is forA developer with a local GPU who wants an agent that works offline.
The problemMany coding agents assume a cloud provider and make local models awkward to wire in.
The outcomeMaki talking to a local Ollama or llama.cpp endpoint through a plain environment variable.
What you need first
- Linux, macOS or Windows
- An API key or sign-in for a supported provider, or a local Ollama or llama.cpp server
- Optional: rtk installed for smaller bash output
Steps
- Install Maki with the script or Homebrew
- For Ollama, set OLLAMA_HOST, for example http://localhost:11434
- For llama.cpp, set LLAMA_CPP_HOST, for example http://localhost:8080
- Start Maki, pick the local model, and use the task tool's weak, medium or strong choice for subagents
Prompt for your agent
Set up tontinton/maki to use my local <Ollama or llama.cpp> server at <host>, pick the best coding model I have, and run a small refactor in <repo> to test it.
Caveat. This card is written from the tontinton/maki README, not from a test run by Deduped. Before you start, make sure you have Linux, macOS or Windows. Check the repo for newer instructions before you rely on these steps.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for tontinton/maki and "Run a coding agent on your own Ollama or llama.cpp server"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer