What it is
Aider is a command-line tool for pair programming with a language model. You add files to the chat, describe a change, and it edits them and makes a git commit for each change. It builds a "repo map" of your codebase so the model understands how files relate. Development has slowed sharply: the latest package, aider-chat 0.86.2, was published on 12 February 2026, and the last commit to the main branch was on 22 May 2026.
Paul Gauthier started it in 2023. It works with almost any model, including local ones, and its public leaderboards of how well models edit code, last updated in November 2025, are still widely cited.
Our review
Aider is in a holding pattern. The last tagged GitHub release is v0.86.0 from August 2025; PyPI has a later point release, 0.86.2, from 12 February 2026. The final commits on the main branch, in May 2026, added newer Claude model names and bash support for the repo map, and nothing has landed since. The repository has about 49,700 stars and 1,938 open issues, 174 of them opened since June.
What it does, it still does well. Every change is a git commit you can review or undo, and the repo map, built with tree-sitter, keeps the model's view of the codebase small. Its own polyglot leaderboard, 225 Exercism exercises across six languages, shows how far cost varies by model: GPT-5 on high reasoning solved 88.0% for $29.08, while DeepSeek-V3.2-Exp reached 74.2% for $1.30.
The gaps are the ones newer agents have closed. There is no MCP support and no sub-agents, and the package requires Python 3.10 to 3.12, so it will not install on 3.13. Analytics are opt-in: a random sample of users is asked, and the docs say code, chat messages and keys are never collected.
The leaderboard has outlived the development pace. The November 2025 table is still quoted in model comparisons, while the tool it was built to measure has shipped one point release since.
Where it earns its place
A backend developer refactoring a Python service
Make AI-assisted changes with an audit trail the team can review
Add the relevant files, describe the change and review the commit Aider makes; /undo reverts it. The only cost is model tokens, and the leaderboard suggests cheaper models such as DeepSeek remain usable.
A developer in an air-gapped research lab
Use an AI pair-programmer without any cloud model
Point Aider at a local model; there is no API bill, at the price of lower edit quality than frontier models. Turn analytics off for good with --analytics-disable.
An engineering manager choosing a terminal agent for 15 developers
Standardise on one tool the team can rely on for a year
Aider is free under Apache 2.0, but with no commits since May and nearly 2,000 open issues, the team would own any breakage when providers change APIs. More active agents such as OpenCode carry less of that risk.
Before you commit
- Setup
- `python -m pip install aider-chat` on Python 3.10 to 3.12, set a provider API key and run `aider` inside a git repository.
- Integrations
- OpenAI, Anthropic, Gemini, DeepSeek and local models; git for every change; no MCP support.
- Data and privacy
- Code you add to the chat goes to whichever model provider you configure. Analytics are opt-in and cover model names, token counts and feature use, not code. Apache 2.0.
- Cost at scale
- No paid edition; the cost is model tokens. On the 225-exercise leaderboard, a full run ranged from $1.30 with DeepSeek-V3.2-Exp to $29.08 with GPT-5 on high reasoning.
- Limits
- No commits since 22 May 2026; Python 3.13 unsupported; support for new models depends on contributors adding them.
What's good
- Every edit is a reviewable git commit
- Tree-sitter repo map keeps context small
- Works with almost any model, including local ones
- Analytics are opt-in and exclude code
What's not
- One point release since August 2025; no commits since May 2026
- No MCP or sub-agents
- Python 3.13 not supported
- 1,938 open issues
Best for
Existing users who want a minimal, git-centred pair-programmer in the terminal.
Alternatives
Aider: common questions
- Is Aider worth it?
- Aider's commit-per-edit discipline still holds up, but a tool with one point release in a year is a habit to keep rather than one to start. We rate it 3.6 out of 5.
- How much does Aider cost?
- Free and open source (Apache 2.0). You pay your model provider or run local models.
- What are the downsides of Aider?
- One point release since August 2025; no commits since May 2026; no MCP or sub-agents; python 3.13 not supported; 1,938 open issues.
- Who is Aider best for?
- Existing users who want a minimal, git-centred pair-programmer in the terminal.
- How much does Aider cost for a team or in production?
- No paid edition; the cost is model tokens. On the 225-exercise leaderboard, a full run ranged from $1.30 with DeepSeek-V3.2-Exp to $29.08 with GPT-5 on high reasoning.
- Is Aider safe for company data?
- Code you add to the chat goes to whichever model provider you configure. Analytics are opt-in and cover model names, token counts and feature use, not code. Apache 2.0.
- What are the best Aider alternatives?
- The closest alternatives we have reviewed are OpenCode, OpenAI Codex CLI and Cline.
Answers checked Oct 23, 2026.