Aider (AI coding)
Aider is an open-source, terminal-based AI pair-programming tool that edits files directly in a local git repository, using an LLM of the user's choice to make code changes and automatically committing each change with a generated commit…
Definition
Aider is an open-source, terminal-based AI pair-programming tool that edits files directly in a local git repository, using an LLM of the user's choice to make code changes and automatically committing each change with a generated commit message.
Overview
Aider, created by Paul Gauthier, takes a deliberately different approach from IDE-integrated tools like Cursor or Windsurf: it runs as a command-line application inside an existing git repository rather than as a graphical editor, and it treats git as a first-class part of its workflow rather than an afterthought. When Aider makes a code change, it applies the edit directly to the relevant files and creates a git commit for that change automatically, with an AI-generated commit message, giving each AI-driven edit a clean, reviewable, revertible unit of history. Aider is model-agnostic: rather than shipping with a single fixed model, it connects to whichever LLM the user configures — commonly Claude or GPT-family models via API keys, but also local models through Ollama or other OpenAI-compatible endpoints — which has made it popular in benchmarking discussions comparing how different frontier models perform at real-world code editing tasks. The project maintains its own public leaderboard (the "Aider polyglot benchmark" and related metrics) tracking how various models perform specifically at Aider-style file-editing tasks across multiple programming languages, which has become a widely cited reference point in the broader conversation about which LLM is "best at coding." Aider builds a repository map using tree-sitter-based code analysis to give the LLM structural context about the codebase — function and class signatures across files — without needing to paste entire files into every prompt, which helps it work effectively even on codebases too large to fit fully in context. It supports both a "chat" mode for asking questions and a direct "edit" mode where instructions produce immediate file changes, along with voice input, image input for describing UI changes, and integration with linters/test runners so it can automatically fix errors it introduces. Because it's open source, free to use (aside from the API costs of whichever model is configured), and terminal-native, Aider has a strong following among developers who prefer working in the terminal or want a lightweight, editor-agnostic AI coding tool rather than adopting a full IDE fork.
Key Features
- Terminal-based, editor-agnostic AI pair programming tool
- Directly edits files and auto-commits each change to git with a generated message
- Model-agnostic: works with Claude, GPT, or local models via Ollama and other endpoints
- Maintains a public benchmark leaderboard ranking model performance on coding tasks
- Builds a tree-sitter-based repository map for structural context without full-file pasting
- Supports chat mode and direct file-editing mode
- Voice input and image input (for describing UI changes) supported
- Open source and free, aside from underlying model API costs
Use Cases
Alternatives
Frequently Asked Questions
From the Blog
Vibe Coding: How to Build Faster with AI Without Losing Control
AI coding tools have shifted from autocomplete to full code generation, multi-file refactoring, and autonomous debugging. This guide explains how to use tools like Copilot, Cursor, and Claude Code effectively — including the critical skill of reviewing AI-generated code before shipping it.
Read More AI & TechnologyAI Coding Assistants: How They Work and When to Use Them
AI coding assistants predict and generate code from your context to speed up development. Learn how they really work, their limits, and when to trust them.
Read More AI & TechnologyAI Coding Assistants at Work: Workflow, Review and Limits
AI coding assistants pay off on well-specified, pattern-heavy work and cost you time on novel design. This guide places them precisely in the development loop, gives you a review order that catches machine-specific defects, and names the failure modes — hallucinated APIs, silent context truncation, leaked secrets — you need to watch for.
Read More AI & TechnologyGiving a coding assistant the repository context it needs
Generic suggestions come from generic context. Learn to make a repo self-describing with conventions, commands and constraints an assistant will actually pick up.
Read More