Aider

Aider was one of the first AI coding agents worth using, and for a long time it was the terminal tool other agents were measured against. It's still free, still open source and still works with nearly any model. What has changed is the pace of development, and that's now the most important thing to know before choosing it.

The idea is simple. You run Aider in your project folder, tell it what to change, and it edits the files and commits the result to git with a sensible message. You pay only for the model behind it.

What it costs to run

Aider has no subscription. The bill is whatever your model provider charges for the tokens Aider sends, which depends on the model, the size of your project and how much back-and-forth a task needs. A local model costs nothing beyond your hardware.

Aider's own leaderboard gives a sense of the spread. Its polyglot benchmark runs 225 coding exercises in C++, Go, Java, JavaScript, Python and Rust, and records what each model's run cost:

The top of Aider's leaderboard on October 4, 2026. Its newest entries are from 2025.

These are 2025 models, but the lesson still applies: the model you pick changes the bill more than anything else. Set a spending limit with your provider before long sessions.

Git as the interface

Aider's design choice is that git does the remembering. Every change it makes becomes a commit with a descriptive message, so you can review it with the diff tools you already use and undo it with a single command. Nothing happens that you can't see in your history.

It also builds a map of your repository, a compact summary of files, classes and functions, and sends it with your requests. That's how it finds the right places to edit in a larger codebase without pasting every file into the prompt. After an edit, Aider can run your linter and tests and fix what they flag.

Beyond the basics, you can add images and web pages to the chat, talk to it by voice, or leave comments in your code for it to act on from any editor.

The slowdown

This is the part to weigh before choosing it. Aider's latest release on PyPI, version 0.86.2, came out on February 12, 2026, six months after the one before it. The main branch has gained support for newer models, including Claude 4.5 and 4.6 and several GPT-5 variants, but those changes haven't shipped in a new release. The last commit to the repository was in May 2026, and its homepage and leaderboard still feature 2025 models.

Aider still works, because it can talk to almost any model through its general connection settings. What you lose is the tuning that came with each release, such as recommended edit formats and model aliases for new models. Expect to set more options by hand.

Where it still fits

Aider remains a good choice for developers who live in the terminal, want every AI change recorded in git and like paying model providers directly. It's light, scriptable and free. It's also a sensible fit for running local models, where privacy matters more than having the newest features.

Maintained alternatives

Cline is the closest open-source alternative and is actively developed, with a VS Code extension, a CLI, a desktop app and a $9.99 ClinePass for open models. Claude Code comes with the $20 Claude plan, and Cursor is a full AI editor at $20 a month. GitHub Copilot Pro costs $10, and Continue.dev is another open-source option for editors. If you'd rather pay for polish, Claude Code vs Cursor weighs the two most popular paid options, and the AI coding tools guide covers the rest.

If you already use Aider and it does what you need, there's no urgent reason to leave. If you're choosing a terminal agent today, compare it with an actively maintained one first.