aider

An AI pair programming tool that runs in the terminal, automatically maps codebases, supports over 100 languages and multi-model access, directly commits every change to Git, and is fully traceable throughout the process.

AI Developer Tools#Open Source#Pair Programming#Terminal

Updated: 2026-10-26

What is aider?

Aider is an AI pair programming tool that runs in the terminal. According to official data, it has 44,000 stars on GitHub, 6.8 million installations, and processes approximately 15 billion tokens per week, making it the most popular open-source project in the command-line AI programming category. It can be used to start a new project from scratch or continue development on an existing codebase, and the official refers to this working method as "pair programming with LLMs". Its core feature is full integration with Git: every change automatically generates a commit, with traceable and rollbackable history. This is the key difference between it and web-based chat tools, and it also provides a clear basis for team collaboration and code reviews. For developers pursuing efficiency, it transforms the action of "modifying code" from copy-pasting to a conversation.

Who is it for?

Engineers who are accustomed to command-line workflows and want every change to be traceable; teams that want to use local models and are unwilling to send code to external services; and developers who need AI to understand the global context in large codebases. It is also suitable for teams that wish to integrate AI coding into a standardized Git workflow for easier auditing.

Key features

Codebase Mapping

Automatically generates a structure diagram for the entire repository, enabling the model to understand the context in large projects. According to the official, this helps it work better on larger projects without manually pasting files into conversations.

Git Integration

Automatically commits every change, enabling easy rollback and comparison, and providing a clear record of each AI modification.

Multi-Model Support

Officially recommended to be used with Claude 3.7 Sonnet, DeepSeek R1 and Chat V3, OpenAI o1, o3-mini, GPT-4o, etc. It can also connect to almost all LLMs, including local models.

Wide language support

Supports over 100 programming languages and can be used as an aid in IDEs.

Rich input methods

Supports input via images and web content, voice input for code, can automatically run lint and tests, and directly copy code to web chat tools to continue working.

How to use aider

Its correct usage is "You take charge of quality control, and it handles typing": a clear commit history is its biggest advantage, but it is still recommended to review each change before proceeding.
When performing multi-file refactoring, first use `/add` to clarify the scope of changes, which can significantly improve accuracy.
different models have quite different code styles, so it is recommended to unify the models used within the team.
When you start using it, first pick a small task to try out, get familiar with the instruction set, and then tackle large projects.

Things to know

The effect is strongly related to the selected model. The token costs of using cutting-edge cloud-based models need to be monitored; for those with budget constraints, switching to local models is an option.

Conversation content and code will be sent to the connected model service providers. For sensitive code, it is recommended to only connect to local models or private deployments.

Auto-generated commits may result in many fragmented commits. It is recommended to adjust the commit granularity as needed or regularly organize the commit history.

The terminal interaction method has a learning curve, and users who are not familiar with the command line need an adaptation period; the tool relies on the local environment (Python and Git), and team deployment requires unified environment configuration.