Aider
Popular open-source pair-programming agent in the terminal; edits across a git repo with any frontier model.
Aider is a free, open-source AI pair-programming tool that runs in your terminal, created by Paul Gauthier (now stewarded by Aider AI LLC). You add files to a chat, describe changes in natural language, and Aider edits the actual files in your local Git repo and auto-commits each change with a descriptive message. It is model-agnostic and git-native, making it a favorite among developers who want a transparent, scriptable coding assistant without subscription lock-in.
- Built by
- open-source (Paul Gauthier)
- Category
- terminal-native AI pair programmer (CLI coding agent)
- Open source
- Yes
- Models
- Model-agnostic. Works with virtually any LLM via API: OpenAI (GPT-5/5.1/5.2, GPT-5-pro, o3, GPT-4o), Anthropic Claude (Opus, Sonnet, Haiku — including 4.x generations), Google Gemini (2.5 Pro/Flash, Gemini 3 preview), DeepSeek, Mistral, and others. Supports local/self-hosted models through Ollama and LM Studio, plus aggregators like OpenRouter.
Available on
Pricing
The tool itself is completely free and open-source (Apache 2.0) — there is no Aider subscription. Your only cost is the API usage of whatever LLM you connect (OpenAI, Anthropic, Google, DeepSeek, etc.); typical heavy users spend roughly $20-$100/month in model API fees, while running fully local models via Ollama/LM Studio costs nothing beyond hardware.
Benchmarks
- Aider Polyglot (Claude Opus 4.5, Anthropic-reported)89.4%
- Aider Polyglot (GPT-5, high reasoning)88.0%
- SWE-bench Lite (Aider as agent, May 2024 SOTA at the time)26.3%
Key capabilities
- ▸Git-native workflow: every AI edit becomes a clean Git commit with a descriptive message, so you can review, diff, or revert with standard Git
- ▸Architect/editor mode: a strong reasoning model plans the change while a cheaper editor model emits the precise file diffs, improving quality and cutting cost
- ▸Tree-sitter repo map: builds a concise map of the whole repository so Aider understands large codebases without loading every file into context
- ▸Multi-modal and multi-input: add images, paste URLs for Aider to read, and code by voice; in-code 'AI!' / 'AI?' comments trigger edits or answers
- ▸/ask and /architect modes for planning and Q&A without immediately editing files
- ▸Broad language support (100+ languages) and works with any API-based or local LLM
Strengths
- +Truly model-agnostic and free — no lock-in, you choose the model and only pay provider API fees (or run local models for free)
- +Git-native and transparent: every change is a reviewable commit, which fits real engineering workflows and makes rollbacks trivial
- +Mature, reliable and well-tested with a large community (~41.6k GitHub stars, 5M+ PyPI installs); strong on multi-file refactors
- +Architect mode lets you pair a top reasoning model with a cheap editor model for high quality at lower cost
Limitations
- −Terminal-only and conversational — no graphical IDE UI, inline autocomplete, or point-and-click experience; steeper learning curve than Cursor/Copilot
- −Bring-your-own-API-key setup required, and unpredictable per-task API costs can add up for heavy use
- −Single-threaded and less autonomous than newer multi-agent/background SWE agents; you drive each step rather than handing off whole tasks
Best for
Developers who live in the terminal and want a transparent, git-centric, model-agnostic AI pair programmer with no subscription lock-in.
What's new
As of 2026 the latest stable release is v0.86.2 (published Feb 12, 2026); recent updates added support for the newest frontier models including OpenAI GPT-5.x / GPT-5-pro, Claude 4.x (Opus/Sonnet/Haiku), and Gemini 3 preview models, with improved architect-mode auto-accept. Aider also famously writes a large share of its own new code each release (~70%+).
Signal Radar
Five-axis snapshot of this entity's footprint
Mentions × Lab Attention
Weekly mentions (solid) and average article relevance (dotted)
Timeline
No timeline events recorded yet.
Predictions
No predictions linked to this entity.
AI Discoveries
No AI agent discoveries for this entity.
Sentiment History
| Week | Avg Sentiment | Mentions |
|---|---|---|
| 2026-W28 | 0.40 | 1 |