What is Metis and how does it elevate LLM coding performance?

Metis is an open-source coding agent harness that wraps around your AI models. Unlike chatbots that output unverified code, Metis elevates model performance through a closed-loop verification cycle: running tests, catching errors, self-healing syntax bugs, and retaining cross-session architectural memory.

Is Metis a free and open-source alternative to Claude Code and Cursor?

Yes. Metis is 100% open-source under the MIT license. You can connect your own API keys (Anthropic, DeepSeek, OpenAI) or run completely offline with local models (Ollama) with zero subscription fees.

Which operating systems and models are supported by Metis?

Metis runs natively on macOS (Apple Silicon and Intel), Windows (native executable and WSL), and Linux. Supported models include Claude 3.7 Sonnet, DeepSeek V3/R1, GPT-4o, o3-mini, and any local GGUF model via Ollama.

Metis

Elevate your model's coding performance. Turn LLM generation into reliable engineering.

The open-source autonomous coding harness built for peak model performance.

Download for macOS

GitHub

npm i -g @wholiver_hu/metis

Metis

Open-Source Coding Agent Harness Built to Maximize LLM Performance