Aider
Terminal coding agent that edits files via your local model. Git-aware, surgical, fast.
Editorial verdict: “Best terminal-native coding agent for local models. Qwen 2.5 Coder 32B is its sweet spot.”
Compatibility at a glance
Which runtime + OS combos this app works against. Source of truth for "will it run on my setup?"
What it is
Aider is for developers who work in the terminal and want a coding agent that respects their existing workflow. It connects to any OpenAI-compatible endpoint—Ollama, llama.cpp, or a remote API—so you can run it fully offline with a local model. The tool’s core strength is its surgical edit format: it produces small, targeted diffs that apply cleanly even with mid-tier models like Qwen 2.5 Coder 32B, and every change is automatically committed as a git commit you can revert. This makes it reliable for iterative refactoring, but the tradeoff is a steeper learning curve compared to IDE-integrated agents. You’ll need at least 24 GB of VRAM to run the recommended 32B-class model for production-quality edits, and the terminal-only interface means no graphical diff review.
✓ Strengths
- +Terminal-native — no IDE lock-in
- +Git-aware: every edit is a commit you can revert
- +Excellent edit-format that minimizes apply failures
△ Caveats
- −Steep learning curve vs IDE-integrated agents
- −Needs 32B-class model for production-quality edits
About the Coding agent category
Editor-integrated or CLI agent that edits code via your model.
Best minimal-surface Copilot-replacement that's been Ollama-native since day one.
Best self-hosted server for teams. SSO + audit logs make it the IT-friendly pick.
Best IDE-integrated agent that fully respects 'all local' as a first-class option.
Best Copilot replacement that defaults to local. Configurable; pair with Qwen 2.5 Coder.
Where to go from here
Pre-filled with this app's recommended use case + budget tier. Get the full rig + runtime + model picks.
The full directory — filter by category, runtime, OS, privacy posture, or VRAM.
What this app talks to: Ollama, vLLM, llama.cpp, MLX, LM Studio. The upstream layer.
Did this app work for you on a specific rig? Submit the benchmark — it powers the model + hardware pages.