Aider
An AI pair programmer in your terminal that writes code right in your project.
Aider reads your Git project, edits files, and makes clean commits. It works with local models via Ollama, so it's free.
What is Aider?
Aider is an AI pair programmer that runs in the terminal. The terminal is the text window where you type commands. You start Aider in your project folder and write, in German or English, what you want to change. Aider sends your request to a language model (an AI model that generates text and code) and then edits the relevant files directly in your project.
How does it work?
Aider works with Git. Git is a program that tracks every change to your files. Aider reads your Git project and builds a kind of map of your entire codebase. The map helps the model find its way around even in larger projects.
After every change, Aider automatically makes a commit, meaning a saved snapshot, with a sensible description. With the usual Git tools, you can view, compare, and undo the AI's changes.
Free with local models
Aider can connect to almost any language model. The developers name Claude 3.7 Sonnet, DeepSeek R1 and V3, and models from OpenAI as particularly well suited, among others. For these cloud models, you usually need an API key (a kind of access code) from the provider.
It becomes free with local models, for example via Ollama. The model then runs on your own computer. As rules of thumb: qwen3-coder:30b needs about 20 GB (V)RAM, meaning system memory or video memory. On weaker hardware, you can try qwen2.5-coder:7b.
More features
- Over 100 programming languages, such as Python, JavaScript, Rust, Go, PHP, HTML, and CSS.
- Use from your development environment (IDE): you write a request as a comment in the code, and Aider gets to work.
- Images and web pages can be pulled into the chat as context, such as screenshots or documentation.
- Voice input: you talk to Aider about your code.
- Linting and tests: Aider checks the code after changes and can fix any errors it finds on its own. Linting is an automatic style check.
- Copy/paste to web chat: you can also use models that you can only reach in the browser.
Limits
Quality depends heavily on the model. Small local models often deliver weaker results than large cloud models. The 30b model also demands a powerful computer. Aider is no substitute for understanding your own code: you should review changes. Without a Git project, the approach doesn't fit well.
What you need
A Git project, a terminal, Python for the installation, and a language model. That can be a local model via Ollama. The license is Apache-2.0.
Ideas: What you can do with it
In an empty Git folder, you describe what you need, such as a small script for renaming photos. Aider creates the files and commits. That quickly gets you a first working version.
You give Aider the error message and name the affected file. Aider proposes a fix and applies it. With Git, you see exactly what changed.
You ask Aider to add tests for a function. With the test integration, your test suite runs after every change, and Aider tries to fix any problems it finds.
You add a screenshot or a web page as context and describe the look you want. Aider changes the HTML and CSS in your project accordingly.
In your usual IDE, you write a comment with your request into the code. Aider picks it up and makes the change without you switching windows.
You speak your wishes aloud, for example a new feature for your smart-home script. Aider implements it and creates a commit.
How to try it
Commands from the official guides – just copy them.
Linux
curl -LsSf https://aider.chat/install.sh | sh && source $HOME/.local/bin/env
ollama pull qwen3-coder:30b
export OLLAMA_API_BASE=http://127.0.0.1:11434
cd dein-projekt && aider --model ollama_chat/qwen3-coder:30bmacOS
curl -LsSf https://aider.chat/install.sh | sh && source $HOME/.local/bin/env
ollama pull qwen3-coder:30b
export OLLAMA_API_BASE=http://127.0.0.1:11434
cd dein-projekt && aider --model ollama_chat/qwen3-coder:30bWindows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://aider.chat/install.ps1 | iex"
$env:Path = "$HOME\.local\bin;$env:Path"
ollama pull qwen3-coder:30b
$env:OLLAMA_API_BASE = "http://127.0.0.1:11434"
cd dein-projekt; aider --model ollama_chat/qwen3-coder:30bTips
- Start with a small project or a copy to get a feel for Aider and the model.
- Give Aider only the files that are actually affected. That keeps the request manageable.
- Phrase small, clear tasks instead of one huge request. That's easier to review and undo.
- Use git diff and the commit list to check every AI change. You can undo anything unwanted with Git.
- If you're short on memory, try qwen2.5-coder:7b instead of qwen3-coder:30b.
If something goes wrong
- The model is too big for your computer: qwen3-coder:30b needs about 20 GB (V)RAM. Switch to the smaller qwen2.5-coder:7b.
- Aider can't find a Git project or behaves strangely: In the terminal, change to the project folder that is a Git repository before you start Aider.
- The results are weak or faulty: This is often down to the small local model. Break the task into smaller steps or try a larger model.
- Cloud models report an access error: They need an API key from the provider. Check that you entered it correctly. Local models via Ollama don't need one.
Ask the AI assistant
Questions about Aider, installation or background? Our assistant knows this page and runs on our own machine, not in someone else's cloud.
The Assistant is an AI (Gemma 4) and can make mistakes – please check important details against the source. Questions are used only to generate an answer and are deleted after one hour. Please don't enter any personal data.
Comments
What do you think? Already tried it? Write it here, no sign-up needed.
Comments are public. Links aren't allowed, and we delete abusive posts. You can delete your own comment at any time in this browser. Learn more in the Privacy Policy.