opendev

opendev

MCP.Pizza Chef: opendev-to

This is a standalone coding agent that you install and run directly from your computer's terminal, rather than a plugin you add to apps like Claude Desktop or Cursor. It launches multiple specialized helpers within one project, letting you assign a different AI model to each — one for writing code, another for reasoning through tricky problems, and another for summarizing progress. You will need your own key from whichever AI provider you choose, such as Claude, OpenAI, or Google's Gemini. A companion web dashboard lets you start a coding task from your phone and check results later.

Coding

Use This MCP client To

Fix bugs in my code while I step away Refactor a large project across many files at once Get a finished set of code changes ready to review Explore my whole codebase and summarize what each part does Run several coding tasks side by side instead of one at a time Check on a coding task from my phone Assign a different AI model to planning versus writing code

README

OpenDev Logo

Open-source AI coding agent that spawns parallel agents, each bound to the LLM of your choice.

GitHub Release Downloads crates.io License: MIT Rust CI Technical Report

Website and documentation coming soon!

OpenDev Demo


Introduction

OpenDev is an open-source, terminal-native coding agent built as a compound AI system. Instead of a single monolithic LLM, it uses a structured ensemble of agents and workflows -- each independently bound to a user-configured model.

Work is organized into concurrent sessions composed of specialized sub-agents. Each agent executes typed workflows (Execution, Thinking, Compaction) that independently bind to an LLM, enabling fine-grained cost, latency, and capability trade-offs per workflow.

OpenDev automatically routes each phase of work to the right model. Five workflow slots — Normal (execution), Thinking (reasoning), Compact (context summarization), Self-Critique (output verification), and VLM (vision) — each bind independently to any LLM you configure. For example, Claude Opus handles execution, GPT-o3 handles reasoning, and a lightweight Qwen model handles compaction — all routed automatically. Together, these form a compound AI system where multiple models collaborate, each optimized for its role.

OpenDev is written in Rust — it starts in 4.3 ms, uses just 9.4 MB of memory, and ships as a single 18 MB binary. That makes it the fastest and lightest coding agent available today — up to 128x faster startup and 30x less memory than alternatives.

opendev FAQ

Can I use this to fix bugs in my code without babysitting it?
Yes — you can hand it a task, step away, and come back to a finished set of changes ready for you to review.
Can I use this to work on several parts of my project at once?
Yes — it can start a whole group of helpers together, each one exploring or fixing a different part of your project at the same time.
Do I need to know how to use a terminal to use this?
Yes — this tool runs from your computer's terminal, or from its companion web dashboard, so it is built for people who are already comfortable writing code.
Which apps does this work with?
It is a standalone tool you install and run on your own computer rather than a plugin for apps like Claude Desktop or Cursor, though it can connect to AI models from providers such as Claude, OpenAI, and Google.
Do I need an account or key from an AI provider?
Yes — you need your own key from at least one AI provider, such as Claude, OpenAI, or Google, before it can get to work.
How hard is it to set up?
It takes some technical comfort — you install it through your computer's terminal using a tool like Homebrew or a downloadable installer, then add your AI provider's key before you can start.
Can I check on a task from my phone?
Yes — a companion web dashboard is built for watching tasks run remotely, including from a phone.
Can I use more than one AI model at the same time?
Yes — you can assign a different AI model to each kind of work, such as one model for writing code and another for double-checking the results.