MiMo Code

MiMo Code is an open-source AI coding agent for the terminal from Xiaomi’s MiMo team, with more than 13,600 stars on GitHub, built around a simple idea: models and agents should evolve together.

Overview

Most coding agents are a fixed harness wrapped around whichever model is fashionable this month. MiMo Code comes from the team that builds the MiMo models themselves, and it is designed as the other half of that work — an agent harness developed alongside the models, so improvements in each side feed the other. In daily use it is a command-line coding companion: you describe a task in natural language, and the agent explores your codebase, plans multi-step changes, edits files, runs commands and tests, and iterates on the results, all from your terminal. It is built for real software engineering sessions rather than single-file demos, keeping context across a project and using tools the way a developer would. The project is MIT licensed, backed by one of the largest model teams in the industry, and moving fast — the repository sees updates on a near-daily basis in 2026.

Key Features

• Terminal-native agent – a full coding workflow that lives in your CLI, next to your existing tools.
• Multi-step task execution – plans, edits, runs commands, and verifies results across a whole codebase.
• Co-evolved with MiMo models – the harness is developed by the same team that trains the MiMo model family.
• Codebase understanding – explores project structure and dependencies before making changes.
• Tool use built in – file editing, shell execution, and search as native agent capabilities.
• Open and hackable – MIT-licensed source you can inspect, configure, and extend.
• Fast-moving project – frequent releases driven by an active model-and-agent team.

💳 Pricing

MiMo Code is free and open source under the MIT license. You install it locally and connect it to MiMo models; any cost comes only from the model access you choose to use, not from the agent itself.

Who Is It For

Developers who live in the terminal and want an autonomous coding partner for multi-file changes, refactoring, and debugging — and anyone curious how a model team’s own agent harness pushes open coding models further.