Updated on
Aider: reviews and analysis
The veteran free terminal coding agent: you bring the API key.
Affiliate link · no extra cost to you
Our verdict
Aider is this category's open source veteran: a terminal agent that pair-programs with you, makes clean commits and runs on whichever model you choose, paying only the API. No subscription and total control make it the choice for anyone who wants to understand what their tool does. The trade-off is that the polish of commercial products is absent: configuration is on you.
Best for: Developers who want total control and to pay only real API usage.
What changed this week
Sweep of
No substantive change from the previous sweep: what we checked this week confirms what was already there.
- The cecli community fork shipped its 1.0.5 release on 4 August 2026 with daily commits, while Aider-AI/aider still has none since 22 May
- down
- no change
We have tracked it since July 25, 2026: 3 sweeps on record, with the score moving between 3.8 and 3.9. We publish the latest change here; the full series is not published.
What the internet says
There is still not a single commit on the main branch since 22 May 2026, and the pull request queue kept growing to 453 open against 680 closed. Real development now lives in cecli, the community fork, which shipped its 1.0.5 release on 4 August 2026 with daily commits last week. It remains free, Apache-2.0 licensed, with the best git integration in the catalogue.
What the web repeats in favour
- Free and Apache-2.0 licensed: you pay only the model provider you choose, with no in-house subscription and no pricing page to hunt for
- Every change lands as a git commit, so reviewing and reverting what the agent does is trivial with the tools you already use
- It works with Claude, OpenAI models, DeepSeek and local models, with no provider lock-in
What the web repeats against
- Development is still stalled: not a single commit since 22 May 2026, and with 453 open pull requests against 680 closed as of today, the review queue keeps growing instead of shrinking
- The documentation has fallen behind: the official leaderboard says it was last updated on 20 November 2025 and the homepage still recommends 2024 and 2025 models, so its guidance on which model to use no longer holds
- Continuity has moved entirely to cecli, the community fork, which shipped its 1.0.5 release on 4 August 2026 fixing the hot-reload command, with daily commits all week
Sweep sources: GitHub · GitHub · GitHub · Docs · Docs · Docs
Pros / Cons
Pros
- Free and open source: you only pay the API
- Exemplary git integration
- Works with almost any model
Cons
- Requires configuring keys and picking models yourself
- Less agentic autonomy than the commercial leaders
TLDR: Aider is this category’s open source veteran: a terminal agent that pair-programs with you, makes clean commits of every change and runs on whichever model you choose, paying only API consumption. No subscription and total control over cost and behavior. The trade-off is that commercial product polish is absent: configuration is on you.
What Aider is and how it works
Before the word agent became fashionable, Aider was already in terminals doing exactly this: taking natural language instructions, editing the project’s files and leaving every change committed to git with its message. It is an open source project with years of community, and that seniority shows in the sound decisions others later copied.
Its architecture is one of control: you choose the model (frontier ones, cheap ones, a local one) and provide your API key, paying exactly the consumption. There is no subscription or intermediary: the tool is free and the bill belongs to whichever model provider you pick.
Its git integration is its most appreciated signature: every change lands as a separate commit with descriptive messages, making it trivial to undo, review or cherry-pick the agent’s work. It is the discipline other agents recommend and Aider enforces by design.
What it is like day to day
The flow is pair programming: you add the relevant files to the context, describe the change and Aider applies and commits it. The session advances at that conversational rhythm, with git history as a permanent safety net.
Manual context management (you deciding which files it sees) is both its friction and its virtue: it demands more intention than agents that explore alone, and in exchange token consumption is predictable and low, because the model does not wander the repository burning context. For anyone watching their API bill, that difference is hard cash.
The honest comparison with the commercial leaders: on tasks you direct, Aider performs at their level for a fraction of the cost. On pure autonomy (exploring the project, deciding what to touch, chaining long steps unguided), Claude Code and company keep the edge that justifies their subscriptions.
Pricing and plans
The tool is free and open source, with no plans or limits. The cost is the API consumption of whichever model you connect: with a cheap model, cents per session. With frontier ones, the bill depends on your volume, and manual context management keeps it notably contained against exploring agents.
That structure makes it the category’s cheapest route for directed use, and the most transparent: you know exactly what you pay and to whom.
Who it is for (and who it is not for)
Aider is for developers who want total control: of the model, the cost, the context and the history. Also for anyone preferring open tools on principle, and for budgets that want an agent without a subscription.
It is not for anyone seeking the zero-configuration experience of commercial products, nor for anyone wanting to delegate project exploration to the agent: here you are the pilot, and that is the offer.
Aider alternatives
OpenCode is the modern open option with a more polished interface and pluggable providers. Cline brings the same bring-your-own-key philosophy to VS Code. And Claude Code is the jump to commercial autonomy when the budget allows. They are in Aider alternatives and the ranking in the best AI coding tools.
Frequently Asked Questions
How much does Aider cost?
The tool, nothing: it is open source. You pay the API consumption of whichever model you connect, which with economical models is cents per session and with frontier ones depends on your volume. Manual context management keeps the bill contained.
Which models does it support?
Practically any: the big providers’ frontier models, the economical ones and local models. Switching models is switching a flag, letting you tune cost and capability per task.
Why is its git integration so often cited?
Because it commits every change separately with descriptive messages, leaving the agent’s work always reviewable and reversible. It is the safety net that turns trust into verification.
Aider or Claude Code?
Directed and cheap versus autonomous and subscription-based. If you know what you want touched and cost matters, Aider more than delivers. If you delegate whole tasks with exploration included, Claude Code’s subscription buys that autonomy.