AINews
CodingSkillsPromptsLearnCompareGlossaryBy roleTools
All tools
AI coding tools

Aider

By Paul Gauthier

Open-source AI pair-programming tool for the terminal. Works with your local git repo, makes proper commits with diffs, and supports any LLM you bring (Claude, GPT, local).

Visit AiderFree

Overview

Aider is the open-source, terminal-based AI pair programmer for engineers who want full control. It runs in your shell, edits files in place, and commits each change to git as a separate commit — so the whole "AI is messing with my repo" anxiety goes away. You can always git revert. It's the spiritual ancestor to Claude Code and still the most opinionated tool in this space. The model is BYOK — bring your own API key for Anthropic, OpenAI, OpenRouter, DeepSeek, Groq, or whatever else. Pay only the per-token cost from the model provider; Aider itself is free. For developers who already have API access, this is the cheapest way to do real agentic coding. It's also the most flexible: swap models per session, run a local model via Ollama, or chain different models for different tasks. The price you pay is polish. Aider has no GUI, no fancy diff viewer, no "accept this suggestion" button. It's a CLI with a chat REPL. If you love the terminal and you trust git, this is liberating. If you don't, it's punishing. As of 2026 it's still actively maintained and remains the reference open-source implementation that Claude Code, Cline, and others learn from.

Best for

  • CLI-driven coding
  • open-source workflows
  • multi-file edits with git history
  • bring-your-own-key setups

Strengths

  • ✓Free and open source — Apache 2.0, no vendor lock-in, no subscription
  • ✓Auto-commits each change to git — full history, easy revert, no AI-eats-my-repo anxiety
  • ✓BYOK model picker — use any provider, swap per session, run local models
  • ✓Terminal-native and scriptable — fits CI, dotfiles, and remote work cleanly
  • ✓Mature codebase: actively developed for years, with strong patterns the rest of the field has copied

Weaknesses

  • ✗No GUI — review-and-accept is chat-based, not diff-pane based
  • ✗You manage costs yourself — easy to surprise yourself with a $40 Claude Opus session
  • ✗Less polished agent loop than Claude Code or Cursor for very long tasks
  • ✗Smaller commercial support story — community-driven, no paid SLA

Pricing

Aider itself

Free

Open source under Apache 2.0. No subscription, no usage caps, no auth wall. Install via pip or uv. The tool is yours.

Model costs (BYOK)

Per-token

You bring an API key. Anthropic, OpenAI, OpenRouter, DeepSeek, Groq, Gemini — anything with an API. Costs scale with your usage; expect $5-50/mo for serious solo work depending on model choice.

Local models

Free (your hardware)

Run a local model via Ollama, llama.cpp, or LiteLLM. No API costs, but you need GPU memory and you'll trade off model quality.

Use cases

  • Solo developers who already pay per-token

    If you have an Anthropic or OpenAI API key, Aider lets you do real AI coding for cents per session — no subscription on top.

  • Open-source maintainers

    Free tool + BYOK + git-native commits = no licensing friction for OSS contributions and easy auditability of AI-authored changes.

  • Working with local models for privacy

    Aider plus Ollama plus a Llama or Qwen model runs entirely offline. Zero data leaves your machine. Useful in regulated work.

  • Multi-model workflows where you pick per task

    Switch from Claude for thinking to DeepSeek for cheap bulk edits to GPT-5 for tricky code. Aider doesn't lock you to one provider.

  • CI/CD and scripted automation

    Aider's CLI is scriptable. Run it in GitHub Actions for migrations, doc updates, or fix-the-flaky-test bots.

  • Engineers who already love the terminal

    If your daily flow is tmux + vim + git, Aider fits in. No editor switch, no new UI to learn.

When not to use

  • ✗You want a graphical IDE experience — use Cursor or Windsurf
  • ✗You don't want to manage API keys or per-token billing
  • ✗You need enterprise procurement, SSO, and SOC 2 — pick a commercial tool
  • ✗You're not comfortable with git as your safety net

Alternatives

Claude Code

Terminal-based agentic coding assistant that reads, edits, and runs commands across your repo.

Cursor

AI-first code editor forked from VS Code, with deep autocomplete and an in-editor chat that knows your repo.

GitHub Copilot

Inline AI autocomplete for VS Code, JetBrains, and other editors, with a chat panel for code questions.

Windsurf

AI-native IDE with Cascade, an agent that can plan and execute multi-step coding tasks autonomously.

Tabnine

Privacy-first AI code completion that can run locally or in your private cloud, with custom model training on your codebase.

See it compared

Aider vs CursorCline vs Aider

Glossary terms to know

LLM (Large Language Model)AI agentTool use / Function callingToken cost

Related skills

Commit Message WriterPR Description WriterRefactor Planner

Other AI coding tools

Cursor

AI-first code editor forked from VS Code, with deep autocomplete and an in-editor chat that knows your repo.

Claude Code

Terminal-based agentic coding assistant that reads, edits, and runs commands across your repo.

GitHub Copilot

Inline AI autocomplete for VS Code, JetBrains, and other editors, with a chat panel for code questions.

Lovable

Builds full-stack web apps from natural-language prompts, with one-click deploy and Supabase integration.

Bolt.new

Browser-based AI app builder that runs full Node.js projects in WebContainers — no install needed.

v0

Generates production-quality React + Tailwind UI components from natural-language prompts.

Windsurf

AI-native IDE with Cascade, an agent that can plan and execute multi-step coding tasks autonomously.

Replit Agent

Cloud IDE where an AI agent builds, runs, and deploys apps from a single prompt — fully browser-based.