A

Aider

Open-source AI pair programming tool that runs in the terminal, edits local code, and keeps changes inside normal git workflows.

free
Last Updated 7/30/2026

About Aider

Aider is an open-source AI pair programming tool for developers who want an AI coding assistant inside the terminal rather than inside one specific editor. The official site describes it as a way to pair program with LLMs to start a new project or build on an existing codebase. Its strongest fit is repository-aware coding work where the developer wants to choose a model provider, select files, ask for changes, inspect diffs, and keep the result in ordinary git history. Aider maps the repository so the model can understand related symbols, integrates with git so edits are committed and undoable, and can run linting or tests after changes. It also supports workflows beyond a plain command prompt, including AI comments from an IDE, images and web pages as context, local models, OpenAI-compatible APIs, and provider-specific API keys. The software is free and Apache 2.0 licensed, but most serious hosted-model use still depends on external model accounts, API keys, token costs, data-handling terms, and model capability. Aider is a strong fit for terminal-native developers and teams that want local control; it is a weaker fit for teams that need a managed cloud agent, visual IDE-first collaboration, centralized policy controls, or a vendor-provided enterprise support plan. Working in an existing repository? Use [Aider vs Cursor for Git-controlled work](/en/articles/aider-vs-cursor-git-controlled-repository) to compare context selection, dirty-worktree handling, diff review, commits, and recovery.

Best for

Developers who prefer a terminal-first AI coding workflow that works across editors, SSH sessions, and existing git repositories.
Engineering teams that want AI-assisted edits to remain visible as normal diffs, commits, branches, and undoable git history.
Open-source maintainers and independent builders who need a model-flexible assistant instead of a single-vendor IDE extension.
Senior engineers evaluating AI coding tools for repository-aware changes, local workflow control, and provider choice.

Best use cases

Implement a small feature inside an existing git repository, let Aider edit the selected files, then review the generated commit and diff before pushing.

Refactor a multi-file module while using the repository map to give the model context about related classes, functions, and call signatures.

Run a failing test or linter after an AI edit, feed the failure back into the same terminal session, and ask Aider to repair the change.

Work over SSH, tmux, a cloud dev box, or a minimal terminal environment where a full AI IDE extension is unavailable or inconvenient.

Compare coding performance across Claude, OpenAI, Gemini, DeepSeek, OpenRouter, local models, or OpenAI-compatible endpoints without changing the editing workflow.

Compare before choosing

Compare with editor-native tools such as Cursor or GitHub Copilot when inline completions, GUI review, and IDE-integrated chat matter more than terminal portability.

Compare with cloud software-engineering agents when you want background task execution, hosted sandboxes, pull-request automation, or managed review queues.

Compare with general chat tools when the main need is explanation or planning, but choose Aider when edits should be applied directly to local files.

Compare with local-model workflows such as Ollama plus a custom script when privacy and offline operation matter more than Aider's built-in git, repo-map, and command UX.

Features

Terminal-based pair programming

Aider runs from the command line, lets developers name files to add to the chat session, and applies requested code changes directly to local project files.

Model-flexible LLM access

Official docs describe connections to many hosted and local models, including OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, OpenAI-compatible APIs, Ollama, and other LiteLLM-supported providers.

Repository map context

Aider builds a concise map of the git repository, including important classes, functions, types, and call signatures, so the model can reason about code outside the immediate files in chat.

Git-integrated change control

Aider is tightly integrated with git, can commit edits with descriptive messages, supports undo and diff review, and takes care around dirty files before applying its own changes.

Linting and testing loop

Aider can automatically lint and test code after it makes changes, and developers can configure custom lint commands or test commands for project-specific verification.

IDE comment workflow

With watch mode, Aider can read AI instructions written as comments in an editor and respond to change or question triggers while preserving a terminal-driven workflow.

Images and web pages as context

Official usage docs explain that developers can add screenshots, mockups, error images, and URLs to the chat for supported models and documentation-heavy tasks.

Privacy-aware analytics controls

Aider's analytics docs say analytics are opt-in, anonymous, and do not collect code, chat messages, keys, or personal information, with a documented command to disable analytics.

Pros

  • Free Apache 2.0 open-source software with an official GitHub repository and public documentation.
  • Works in terminal environments where editor-specific AI extensions may not be available, including remote shells and lightweight dev boxes.
  • Keeps AI edits inside normal git review habits, including diffs, commits, undo flows, branches, and external git tools.
  • Supports a broad range of model providers and local-model routes instead of locking the workflow to one hosted coding model.
  • Repository maps, linting, testing, image context, web-page context, and IDE comment triggers make it more capable than a simple chat wrapper.

Cons

  • The software is free, but most productive hosted-model use still requires separate API keys, provider accounts, token budgets, and provider-specific data terms.
  • A terminal-first interface can be less approachable for developers who prefer inline IDE completions, visual file trees, and GUI review panels.
  • Aider is not a managed cloud engineering agent; it does not by itself provide hosted sandboxes, queued background tasks, or organization-wide admin controls.
  • Model quality matters: the official docs warn that weaker models may fail to return usable code edits, so results depend heavily on model choice.
  • Teams handling sensitive code still need to evaluate the selected LLM provider, local model setup, analytics settings, and internal policy before adoption.

Pricing Plans

Open-source Aider software

$0No Aider software billing cycle
  • Apache 2.0 licensed open-source project
  • Terminal coding assistant for local git repositories
  • Repository map, git integration, lint/test hooks, and configuration support
  • Install options documented for aider-install, uv, pipx, pip, Docker, Codespaces, and related routes

External model/provider usage

Varies by providerExternal provider terms
  • Most hosted models require an API key or provider account
  • Costs depend on selected model, token volume, rate limits, and provider plan
  • Free provider options and local models may reduce software spend but can have capability or quota limits
  • Provider privacy, retention, regional availability, and support terms should be checked before using private code

* Pricing information is for reference only. Please refer to the latest announcement on Aider's official website.

FAQ

Aider is an open-source AI pair programming tool that runs in the terminal and edits files in a local codebase with help from large language models.
Yes. The official FAQ says Aider is open source and available on GitHub under the Apache 2.0 license. Model usage may still cost money through the selected LLM provider.
Aider can work with many hosted and local models. Official docs mention OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, OpenAI-compatible APIs, Ollama, and other models through LiteLLM.
Most hosted-model workflows require an API key from a provider. The docs also describe environment variables, .env files, YAML config, and command-line switches for managing those keys.
Aider builds a repository map that summarizes important files, classes, functions, types, and call signatures, then includes the most relevant parts in the model context.
Aider works best in a git repo. It can commit AI edits with descriptive messages, show diffs, undo the last change, and keep preexisting dirty-file changes separate from Aider's own edits.
Yes. Aider can automatically lint and test code after edits, and developers can configure custom commands for project-specific linting or testing workflows.
Aider's own analytics are documented as opt-in and anonymous, but code privacy also depends on the selected model provider or local-model setup. Teams should review provider terms and internal policy before sending private code.

Tool Information

DeveloperAider
Release Date4/5/2026
Pricingfree

Related Tags

AI coding assistantTerminal developer toolOpen-source softwareGit workflowDeveloper productivity