Aider

Aider AI (open source) Terminal agents Free way to start

Open-source AI pair programmer in your terminal. Works with almost any LLM via your own API keys, including local models.

Verified against official sources · Last checked: 2026-09-19

Capabilities

Read local project Supported Builds a map of your whole repo.
Edit files Supported Edits files with automatic git commits.
Run commands Supported Run shell commands and tests; aider can fix failing tests.
Image input Supported Add screenshots/images to the chat.
Long autonomous tasks Conditional Pair-programming style — you steer each step rather than fire-and-forget.

Working experience

Code quality Good
Works on its own Limited
Speed Good
Stability Good

Editorial assessment based on official documentation and publicly known behavior — not a field-test score. Sources and check date on this page.

🧪 ResetDog field notes

No completed field test for this tool yet. The capability descriptions on this page come from official documentation.

🎯 Good for

Change pages / UI · Fix a bug · · Continue from a screenshot · Quick code question · Hand off my project

🧭 How you use it

In the terminal
Works on: macOS / Windows / Linux

💰 Cost & quota

Bring your own key · Free / open source

🆓 Free / trial access

Tool is free and open source (Apache-2.0); you pay your model provider per token. Works with local models for a $0 pipeline.

🏷️ Pricing notes

No license fee. Costs = your chosen API usage (e.g. DeepSeek or Gemini Flash are cheap; Claude/OpenAI cost more).

See pricing page ↗

🚀 Setup

Python required: `python -m pip install aider-install && aider-install`, then run `aider` with `--model` and your API key.

⚠️ Main limitations

  • Terminal-only workflow; not an autonomous background agent.
  • You manage API keys and costs yourself.

🔒 Privacy & data

No aider server involved: requests go directly from your machine to the LLM provider you pick.

🧾 What is official vs ours

Facts like pricing, plans, and capability limits come from the official pages linked below, checked on the date shown. Summaries and task-fit notes are written by ResetDog. Anything marked “conditional” or “unknown” was not fully confirmed — treat it as unverified.