Aider
Aider is an open-source, terminal-based AI pair programmer that edits code directly inside a developer's local Git repository. Written in Python and distributed via PyPI under the Apache 2.0 license, Aider is a BYO-LLM tool: the user supplies API keys for hosted models (Anthropic Claude, OpenAI, DeepSeek, Google Gemini, OpenRouter, Mistral, xAI, GROQ, Cohere, GitHub Copilot, Azure OpenAI, Amazon Bedrock, Vertex AI) or points it at local models running through Ollama, LM Studio, or any OpenAI-compatible endpoint. Distinguishing capabilities include a repository-wide codebase map that lets the model reason about files it has not been shown, multi-file architectural edits with diff-based patching, automatic Git commits with conventional commit messages, a lint/test/repair loop, image and URL context, voice-to-code, and an IDE bridge via `--watch-files` that picks up `AI!` / `AI?` comments written from any editor. Aider supports 100+ programming languages via tree-sitter, publishes a public LLM coding leaderboard (the polyglot benchmark across C++, Go, Java, JavaScript, Python, Rust), and has accumulated 45K+ GitHub stars with 4.5K+ forks. The product surface is a rich CLI (40+ in-chat slash commands across four chat modes: code, architect, ask, help) plus a Python module entrypoint (`python -m aider`); there is no hosted SaaS, no Aider-operated REST API, and no app-level rate-limiting or billing — all model inference is delegated to the user's chosen provider, and any pricing, rate limits, and FinOps surface live with that provider.
2 APIs
0 Features
AIAI Pair ProgrammingDeveloper ToolsCLICommand LineCoding AssistantCode GenerationOpen SourcePythonApache 2.0LLMGitBYO LLMTerminalPolyglotTree SitterRepository MapPair Programming
aid: aider
name: Aider
description: >-
Aider is an open-source, terminal-based AI pair programmer that edits code directly inside a developer's local Git
repository. Written in Python and distributed via PyPI under the Apache 2.0 license, Aider is a BYO-LLM tool: the user
supplies API keys for hosted models (Anthropic Claude, OpenAI, DeepSeek, Google Gemini, OpenRouter, Mistral, xAI,
GROQ, Cohere, GitHub Copilot, Azure OpenAI, Amazon Bedrock, Vertex AI) or points it at local models running through
Ollama, LM Studio, or any OpenAI-compatible endpoint. Distinguishing capabilities include a repository-wide codebase
map that lets the model reason about files it has not been shown, multi-file architectural edits with diff-based
patching, automatic Git commits with conventional commit messages, a lint/test/repair loop, image and URL context,
voice-to-code, and an IDE bridge via `--watch-files` that picks up `AI!` / `AI?` comments written from any editor.
Aider supports 100+ programming languages via tree-sitter, publishes a public LLM coding leaderboard (the polyglot
benchmark across C++, Go, Java, JavaScript, Python, Rust), and has accumulated 45K+ GitHub stars with 4.5K+ forks. The
product surface is a rich CLI (40+ in-chat slash commands across four chat modes: code, architect, ask, help) plus a
Python module entrypoint (`python -m aider`); there is no hosted SaaS, no Aider-operated REST API, and no app-level
rate-limiting or billing — all model inference is delegated to the user's chosen provider, and any pricing, rate
limits, and FinOps surface live with that provider.
type: Index
position: Provider
access: Open Source
image: https://kinlane-images.s3.amazonaws.com/shared/apis-json/apis-json-logo.jpg
x-type: opensource
tags:
- AI
- AI Pair Programming
- Developer Tools
- CLI
- Command Line
- Coding Assistant
- Code Generation
- Open Source
- Python
- Apache 2.0
- LLM
- Git
- BYO LLM
- Terminal
- Polyglot
- Tree Sitter
- Repository Map
- Pair Programming
url: https://raw.githubusercontent.com/api-evangelist/aider/refs/heads/main/apis.yml
created: '2026-05-08'
modified: '2026-05-30'
specificationVersion: '0.20'
apis:
- aid: aider:cli
name: Aider CLI
description: >-
Open-source terminal AI pair programmer distributed as the `aider-chat` PyPI package (entrypoint `aider`).
Federates to user-supplied LLM providers (Anthropic, OpenAI, DeepSeek, Google Gemini, OpenRouter, Mistral, xAI,
GROQ, Cohere, GitHub Copilot, Azure, Bedrock, Vertex, or any OpenAI-compatible endpoint including local models via
Ollama and LM Studio). Provides four chat modes (code, architect, ask, help), 40+ in-chat slash commands, a
repository-wide codebase map built with tree-sitter, multi-file diff-based edits in several edit formats (diff,
whole, udiff, editor-diff), Git auto-commit with conventional commit messages, lint/test/repair loop, image and
URL context, voice-to-code via the `/voice` command, and an IDE bridge via `--watch-files` that picks up `AI!` /
`AI?` comments written from any editor. Configuration via command-line flags, environment variables (`AIDER_*`),
`.env`, or `.aider.conf.yml`.
image: https://kinlane-images.s3.amazonaws.com/shared/apis-json/apis-json-logo.jpg
humanURL: https://aider.chat/
baseURL: https://aider.chat
version: 0.86.1
tags:
- CLI
- Open Source
- AI Pair Programming
- BYO LLM
- Python
- Tree Sitter
- Repository Map
properties:
- type: Documentation
url: https://aider.chat/docs/
- type: GitHub
url: https://github.com/Aider-AI/aider
- type: PyPI
url: https://pypi.org/project/aider-chat/
- type: Installation
url: https://aider.chat/docs/install.html
- type: UsageGuide
url: https://aider.chat/docs/usage.html
- type: Commands
url: https://aider.chat/docs/usage/commands.html
- type: ChatModes
url: https://aider.chat/docs/usage/modes.html
- type: Configuration
url: https://aider.chat/docs/config.html
- type: ConfigOptions
url: https://aider.chat/docs/config/options.html
- type: ConfigYAML
url: https://aider.chat/docs/config/aider_conf.html
- type: ConfigEnv
url: https://aider.chat/docs/config/dotenv.html
- type: APIKeys
url: https://aider.chat/docs/config/api-keys.html
- type: SupportedModels
url: https://aider.chat/docs/llms.html
- type: ModelWarnings
url: https://aider.chat/docs/llms/warnings.html
- type: EditFormats
url: https://aider.chat/docs/more/edit-formats.html
- type: Leaderboard
url: https://aider.chat/docs/leaderboards/
- type: ChangeLog
url: https://aider.chat/HISTORY.html
- type: License
url: https://github.com/Aider-AI/aider/blob/main/LICENSE.txt
- type: SourceCode
url: https://github.com/Aider-AI/aider
- type: IssueTracker
url: https://github.com/Aider-AI/aider/issues
- type: Tutorials
url: https://aider.chat/docs/usage/tutorials.html
- type: WatchFilesGuide
url: https://aider.chat/docs/usage/watch.html
- type: VoiceGuide
url: https://aider.chat/docs/usage/voice.html
- type: ConventionsGuide
url: https://aider.chat/docs/usage/conventions.html
- type: OpenAPI
url: openapi/aider-cli-openapi.yml
- aid: aider:python-module
name: Aider Python Module
description: >-
Aider exposes a Python module entrypoint (`python -m aider`) and an importable package (`aider.main.main`) so
Aider can be embedded in Python scripts, CI workflows, and other agents. The same flags and configuration that
drive the CLI apply to module invocation. There is no documented stable Python SDK contract beyond `main()` — the
internal `Coder`, `Model`, `IO`, and `Commands` classes are public but versioned with the CLI and may change
between releases.
image: https://kinlane-images.s3.amazonaws.com/shared/apis-json/apis-json-logo.jpg
humanURL: https://github.com/Aider-AI/aider/blob/main/aider/main.py
baseURL: https://github.com/Aider-AI/aider
tags:
- Python
- Library
- Embedding
- Open Source
properties:
- type: SourceCode
url: https://github.com/Aider-AI/aider/blob/main/aider/main.py
- type: PyPI
url: https://pypi.org/project/aider-chat/
- type: Documentation
url: https://aider.chat/docs/scripting.html
- type: Scripting
url: https://aider.chat/docs/scripting.html
common:
- type: Website
url: https://aider.chat/
- type: Documentation
url: https://aider.chat/docs/
- type: GitHub
url: https://github.com/Aider-AI/aider
- type: GitHubOrg
url: https://github.com/Aider-AI
- type: Issues
url: https://github.com/Aider-AI/aider/issues
- type: Discussions
url: https://github.com/Aider-AI/aider/discussions
- type: PyPI
url: https://pypi.org/project/aider-chat/
- type: Installation
url: https://aider.chat/docs/install.html
- type: UsageGuide
url: https://aider.chat/docs/usage.html
- type: Configuration
url: https://aider.chat/docs/config.html
- type: SupportedModels
url: https://aider.chat/docs/llms.html
- type: Leaderboard
url: https://aider.chat/docs/leaderboards/
- type: Blog
url: https://aider.chat/blog/
- type: ChangeLog
url: https://aider.chat/HISTORY.html
- type: License
url: https://github.com/Aider-AI/aider/blob/main/LICENSE.txt
- type: Discord
url: https://discord.gg/Tv2uQnR88V
- type: Twitter
url: https://twitter.com/paulgauthier
- type: Conventions
url: https://github.com/Aider-AI/conventions
- type: Benchmarks
url: https://github.com/Aider-AI/polyglot-benchmark
- type: Tools
title: aider-install Bootstrapper
url: https://github.com/Aider-AI/aider-install
- type: Tools
title: grep-ast Repository Map Helper
url: https://github.com/Aider-AI/grep-ast
- type: Tools
title: Polyglot Benchmark Suite
url: https://github.com/Aider-AI/polyglot-benchmark
- type: Tools
title: Refactor Benchmark Suite
url: https://github.com/Aider-AI/refactor-benchmark
- type: Tools
title: SWE-Bench Harness
url: https://github.com/Aider-AI/aider-swe-bench
- type: Tools
title: MCP Server (Community — disler/aider-mcp-server)
url: https://github.com/disler/aider-mcp-server
- type: Tools
title: MCP Server (Community — sengokudaikon/aider-mcp-server)
url: https://github.com/sengokudaikon/aider-mcp-server
- type: Tools
title: MCP Package Manager for Aider (mcpm-aider)
url: https://github.com/lutzleonhardt/mcpm-aider
- type: Integrations
title: NeoVim Plugin (aider.nvim)
url: https://github.com/joshuavial/aider.nvim
- type: Rules
url: rules/aider-rules.yml
- type: Vocabulary
url: vocabulary/aider-vocabulary.yml
- type: RateLimits
url: rate-limits/aider-rate-limits.yml
- type: JSONLD
url: json-ld/aider-cli-context.jsonld
features:
- name: Repository Map
description: >-
Aider builds a tree-sitter-powered map of the entire repository so the LLM can reason about files that haven't
been explicitly added to the chat. The map is automatically refreshed and can be inspected via `/map` or rebuilt
via `/map-refresh`.
- name: Architect / Editor Split
description: >-
Architect mode pairs a reasoning-strong model (planning) with a separate editor model (file edits). This separates
the cost of thinking from the cost of writing diffs and lets weaker editor models execute plans from stronger
architects.
- name: Watch Files Mode
description: >-
`--watch-files` lets aider sit in a terminal alongside any editor; it reacts to `AI!` and `AI?` comments authored
from VS Code, Cursor, JetBrains, Vim, or any other editor.
- name: Voice to Code
description: >-
`/voice` records microphone input, transcribes it (Whisper-style), and feeds it into the chat as a prompt.
Optional PortAudio install required on Mac/Linux.
- name: Auto Git Commits
description: >-
Every AI-applied change is automatically committed with a conventional commit message. `/undo` rolls back the last
commit; `/diff` shows what changed since the last message.
- name: Lint and Test Repair Loop
description: >-
`--auto-lint` and `--auto-test` re-run linters and tests after every edit; failures are fed back into the LLM so
it can self-repair until the suite passes.
- name: Polyglot Benchmark
description: >-
Aider publishes a public leaderboard scoring frontier LLMs against 225 Exercism exercises across C++, Go, Java,
JavaScript, Python, and Rust — separating edit-format accuracy from raw correctness.
- name: BYO LLM
description: >-
No vendor lock-in. The user supplies the API key for any of 15+ supported providers (or points at a local Ollama /
LM Studio endpoint). Pricing, rate limits, and billing live entirely with the upstream provider.
- name: Multiple Edit Formats
description: >-
Aider supports `diff`, `whole`, `udiff`, and `editor-diff` edit formats so the LLM can apply changes in whatever
shape it executes most reliably. Edit format is auto-selected per model and can be overridden via `--edit-format`.
useCases:
- name: Multi-File Refactor
description: >-
Use architect mode with a reasoning model to plan a cross-cutting refactor, then have aider apply the edits across
dozens of files in one pass, with each edit auto-committed.
- name: Test-Driven Bug Fixing
description: >-
`/test` runs the suite, the failure goes back to the model, the model proposes a fix, `/test` re-runs until green.
The whole loop happens without leaving the terminal.
- name: Editor-Bridged Pair Programming
description: >-
Run `aider --watch-files` in one pane and any editor in another; drop `# AI! make this idempotent` into the source
file and aider picks it up, edits, and commits.
- name: Local Model Coding
description: >-
Point aider at an OpenAI-compatible Ollama or LM Studio endpoint to get a fully offline coding workflow with no
data ever leaving the developer's machine.
- name: Benchmark Authoring
description: >-
Use the polyglot, refactor, and SWE-Bench harnesses (in the Aider-AI GitHub org) to benchmark a new model or new
prompt strategy against a reproducible, published scoreboard.
integrations:
- name: Anthropic Claude
description: >-
First-class support for Claude Sonnet, Opus, and Haiku via Anthropic direct API or Bedrock / Vertex. `--model
claude-...` or `ANTHROPIC_API_KEY` in the environment.
- name: OpenAI
description: >-
Support for GPT-4.1, o3, o4-mini, GPT-5 family, and Responses API models (o1-pro, o3-pro). `--model gpt-...` or
`OPENAI_API_KEY`.
- name: Google Gemini
description: Gemini 2.5 Pro and Flash via direct API or Vertex. Free tier available through `gemini-2.5-pro-exp`.
- name: DeepSeek
description: DeepSeek R1 (reasoner) and V3 via the DeepSeek API or third-party providers.
- name: OpenRouter
description: Unified gateway to dozens of frontier and open models, with daily free quota.
- name: xAI Grok
description: Grok-4 and earlier Grok models via xAI API.
- name: Mistral
description: Mistral hosted models via Mistral La Plateforme.
- name: Ollama
description: Local model execution for fully offline coding workflows.
- name: LM Studio
description: Local OpenAI-compatible endpoint for self-hosted models.
- name: Azure OpenAI
description: Enterprise OpenAI deployments via Azure with custom endpoint support.
- name: Amazon Bedrock
description: Claude and other Bedrock-hosted models via AWS credentials.
- name: Vertex AI
description: Gemini and partner models via Google Cloud Vertex AI.
- name: GROQ
description: Fast inference for Llama and Mixtral via GROQ Cloud.
- name: GitHub Copilot
description: Aider can route through a GitHub Copilot subscription as the model backend.
- name: Cohere
description: Cohere Command-family models for chat and code completion.
- name: Git
description: >-
Aider requires a Git repo by default; every applied edit is committed automatically. Configurable via `--git` /
`--auto-commits` / `--dirty-commits`.
- name: NeoVim (aider.nvim)
description: Community NeoVim plugin pairs aider with an in-editor experience.
- name: VS Code Terminal
description: >-
No official extension; runs natively in VS Code's integrated terminal. `--watch-files` makes the IDE pairing work
via AI! / AI? comments.
solutions:
- name: Offline Coding Assistant
description: >-
Run aider against a local Ollama or LM Studio server to get AI-assisted refactoring, test repair, and
architectural edits with no data ever leaving the developer's machine.
- name: Cross-Model Benchmarking
description: >-
Use Aider's polyglot benchmark harness to score any new LLM against a reproducible 225-task suite across six
languages, with per-language and edit-format-accuracy breakdowns.
- name: Editor-Native AI Workflow
description: >-
Pair `--watch-files` with VS Code, JetBrains, NeoVim, or any editor to drive aider from inline `AI!` / `AI?`
comments — no extension required.
- name: MCP-Wrapped Aider
description: >-
Community MCP servers (`disler/aider-mcp-server`, `sengokudaikon/aider-mcp-server`) expose aider AS an MCP tool,
letting Claude Code, Cursor, or any MCP client offload editing work to aider.
maintainers:
- FN: Kin Lane
email: kin@apievangelist.com