Aider

Name: Aider
Author: AgentRadar

Verified

Open-source terminal pair-programming agent with git-native workflow

Coding Agents Open Source

8.6 /10

Visit Website GitHub

Released

2023-05

Country

United States

API

N/A

Self-Host

Yes

GitHub Stars

46,000

Funding

Self-funded

Last Updated

2026-06

About Aider

Aider is a terminal-based pair-programming agent that works directly in your git repo alongside any LLM you bring — Claude, GPT, DeepSeek, o-series models, or local models via Ollama. It builds a repo map that gives the model structural awareness of your codebase without dumping every file into context, which keeps token usage efficient even on large projects. Every change is automatically committed to git with a descriptive message, so your history stays clean and each step is reversible. Aider supports over 100 programming languages and can lint and test after each edit to catch regressions before they land. It also accepts image and web content as input, supports voice-to-code, and lets you paste conversations from web-based LLM chats directly into the terminal. Notably, Aider writes 60-90% of its own code — a strong signal of the tool's maturity and self-improving nature.

Verdict

The most flexible terminal pair-programming agent for developers who want model choice, strong git workflow, and efficient token usage on large codebases. Aider's repo map and auto-commit discipline make it the terminal agent of choice for experienced developers who value clean history and cost control over IDE polish. Its BYOK approach means you are never locked into a single model provider.

Features

Automatic git commits with descriptive messages

Repo map for efficient codebase context

100+ programming language support

Bring-your-own-key for Claude, GPT, DeepSeek, local models

Automatic linting and testing after edits

Voice-to-code and image/web context input

Paste from web LLM chats to resume in terminal

Works with o-series reasoning models

Detailed Ratings

Ease of Use

7.5

Value for Money

9.5

Features

8.5

Support

8.0

Performance

9.0

Overall Rating

8.6 /10

Pros & Cons

Pros

Best-in-class git workflow — every change auto-committed with clear history
Repo map provides codebase awareness without excessive token usage
Model-agnostic BYOK: Claude, GPT, DeepSeek, local models — full control over cost and capability
Proven quality — Aider writes 60-90% of its own code
Open-source Apache 2.0 with no vendor lock-in

Cons

Requires a git-initialized project — not usable in unversioned directories
Terminal-only with no IDE or GUI integration
Users manage their own API costs — BYOK means variable spend depending on usage

Use Cases

Terminal pair programming with strong git hygieneCost-controlled coding with BYOK model selectionLarge codebase work with efficient repo map contextResuming web-based LLM conversations in a terminal environmentVoice-driven or image-guided coding workflows

Who Is It For?

Experienced developers who want a model-agnostic terminal pair-programming agent with strong git workflow and efficient codebase context

#cli#terminal#pair-programming#open-source#git-native#byok

Pricing Plans

Open Source

Free

Full CLI with no feature gates
Bring your own LLM API key
All model providers supported
Apache 2.0 license

LLM API Costs

Usage-based /BYOK

Pay your LLM provider directly (Anthropic, OpenAI, Google, DeepSeek)
Typical spend $10-50/month for daily use
Local models via Ollama are free

Frequently Asked Questions

Does Aider require an Anthropic or OpenAI subscription?

No. Aider is open-source and free — you bring your own API key for any supported provider (Anthropic, OpenAI, Google, DeepSeek, Ollama for local models). You pay the provider directly, and Aider does not take a cut.

What is the repo map and how does it work?

Aider builds a tree-structured map of your repository that summarizes each file's purpose, classes, and functions. This gives the LLM structural understanding of the codebase without loading full file contents into context, keeping token usage low on large projects.

How is Aider different from Claude Code or Codex CLI?

Aider is model-agnostic — it works with any LLM you bring, not just one provider's models. Its signature strength is git-native workflow: every edit is automatically committed. Claude Code and Codex CLI are first-party tools tied to Anthropic and OpenAI respectively, with broader IDE integration.

What does 'Aider writes 60-90% of its own code' mean?

Paul Gauthier uses Aider itself to develop Aider. The tool is its own primary user, which serves as continuous integration testing of its own capabilities — the strongest possible dogfooding signal.