🇺🇸 USA · Aider (open-source CLI)

Status: 🟩 COMPLETE 🟦 LIVING Last updated: 2026-06-26 Plain-English tagline: The original open-source AI coding CLI. Run in your terminal, BYO API key, edit files via natural-language conversation. Predates Claude Code; still actively maintained.

Front-matter facts

Field	Value
Vendor	Paul Gauthier (USA) — primary maintainer; broad open-source community
Country / origin	🇺🇸 USA (open-source)
Recommended for Australian users?	✅ Yes — fully open-source, no region restrictions
Privacy summary	Aider itself sends nothing to a central server; your data goes to your chosen model API
Free tier	Aider itself is free; API costs apply via your chosen model provider
Paid tiers	None for Aider; model API costs apply
First released	May 2023
Last reviewed	2026-06-26
Official site	https://aider.chat

What it is

Aider is a command-line AI coding assistant — one of the first open-source tools to bring “chat-with-your-codebase” workflow to the terminal. It pre-dates Claude Code by over a year and remains one of the most respected open-source options.

You install Aider, point it at a git repo, configure a model API key (Claude, GPT, Gemini, Mistral, local via Ollama, etc.), and have natural-language conversations that result in file edits. Aider handles the git plumbing — every change is committed with a descriptive message, so you have full undo / review via standard git.

Key features:

Open-source (Apache 2.0) — full source on GitHub
Model-agnostic — works with Claude, GPT, Gemini, Mistral, DeepSeek (via OpenRouter), local models
Git-native — every change is a commit; easy to review and undo
Terminal-based — works in any terminal; no IDE required
Map-aware — Aider builds a repo map to help models understand large codebases efficiently
Voice mode — you can dictate prompts
Long history — Paul Gauthier (the maintainer) publishes regular benchmarks comparing Aider’s performance across models

What you’d use it for

Terminal-first developer workflow with AI assistance
You want open-source that you can audit / modify
Polyglot environments (any language Aider can see the files of)
CI / automation scripts that need AI-driven edits
Server / remote work without installing a full IDE
Trying many models — Aider’s neutral API surface makes switching trivial
Educational / academic settings where vendor-neutral tooling is preferred

How to install + first 5 minutes from Australia

Install:
```
pip install aider-install
aider-install
```
(or use pipx, uv, brew, scoop — see aider.chat/docs/install)
Get an API key from your chosen provider:
- Anthropic: console.anthropic.com
- OpenAI: platform.openai.com
- Google: aistudio.google.com
- Or run local: install Ollama / LM Studio

Set the key:

export ANTHROPIC_API_KEY=sk-ant-...
# or OPENAI_API_KEY, GOOGLE_API_KEY, etc.

Run in a git repo:
```
cd ~/my-project
aider
```
Try it:
- “Add a /health endpoint to the Express server”
- “Refactor the auth middleware to use JWT”
- “Add unit tests for the user service”

Aider proposes diffs, shows them in the terminal, asks for confirmation, then commits.

What it costs

Aider: free (open-source Apache 2.0)

API costs

Anthropic Claude Sonnet ~US $3/$ 15 per million tokens
OpenAI GPT-5 — similar pricing
Google Gemini — varies
Local (Ollama) — free, hardware-limited
Typical Aider session: US$0.50-5 in API costs for moderate work

How it compares to alternatives

Capability	Aider	Claude Code	Cline	Cursor
Open-source	Yes (Apache 2.0)	No	Yes (Apache 2.0)	No
Surface	Terminal CLI	Terminal CLI	VS Code extension	Standalone IDE
BYO API key	Yes (any provider)	API or subscription	Yes	Subscription
Multi-model	Yes (broadest)	Claude family	Multi	Multi
Git integration	Native (commits per change)	Manual	Via VS Code	Via Cursor
Skills / Plugins	Limited	Best (Skills + MCP)	MCP support	Yes
Long maintenance history	Since May 2023	Since 2024	Since 2024	Since 2023
Best for	OSS-purist terminal user	Anthropic-stack daily driver	VS Code stay-in-place	Editor-first

Aider’s distinguishing value: pure open-source + terminal + multi-model + git-native in one package. For developers who value that combination, no commercial alternative matches.

Privacy / data handling

Aider itself sends nothing to a central server — all data flows to your chosen model API
Privacy posture = your chosen model provider’s posture
Local models (Ollama) keep everything on your machine
No telemetry to Aider maintainers by default

Recent changes

2026: Continued improvements to repo-mapping; expanded local-model support
2025: Benchmark publication regularly; Aider became reference for “which model is best at coding tasks”
May 2023: Initial release

Gotchas

You need an API key — Aider doesn’t provide one
Direct API billing is metered — heavy use can cost more than a subscription. Set provider-side spend limits.
Git commits are real — Aider commits every change. Easy to undo (git reset), but worth knowing it’s modifying your git history actively
Repo map for large codebases — Aider builds an internal map; very large repos (millions of LOC) can take time on first run
No MCP support in core Aider as of mid-2026 — for MCP-heavy workflows, Claude Code is the choice
Doesn’t run servers or test code as autonomously as Claude Code — Aider focuses on edit-then-commit; agentic execution requires more configuration
Voice mode is excellent on Mac (uses macOS speech); other platforms vary

Tech & AI, Explained

Explorer

aider