Overview
Aider is an open-source terminal AI pair programming tool focused on collaborating with developers inside real code repositories. Its most praised trait is deep Git integration—every AI code change automatically produces a meaningful Git commit, keeping version history clear and traceable.
Aider reaches models through the LiteLLM layer, so its model coverage is very broad, spanning cloud APIs from various vendors as well as local Ollama models. On benchmarks like SWE-bench, it has long sat in the top tier of open-source terminal tools. One thing worth noting: **Aider's maintenance cadence slowed noticeably in 2026**—the latest release v0.86.2 landed on February 12, 2026, the last commit on the main branch was on May 22, 2026, and the recommended model list on its homepage still reflects the 2025 generation. The project remains usable and is licensed under Apache-2.0, but it is worth evaluating update frequency as a separate criterion when choosing a tool.
Key Features
- Smart Git Integration: Every code change automatically generates a well-formed Git commit, keeping version history clear and traceable
- Large Codebase Support: Uses repo map technology to understand the whole project structure, excelling at cross-file refactoring
- Free Model Switching: Connects to mainstream LLMs through LiteLLM, supporting both cloud and local models, with examples including DeepSeek-V4.1-Flash, Claude, GPT-5.x, Qwen3.8, and local Ollama models
- Conversational Programming: Talk naturally in the terminal; describe what you need and it edits the code, with multi-turn iteration support
- Voice Programming: Supports voice input for hands-free programming conversations
- Image and Web Page Support: Supports sending images and web pages into the chat for visual context, screenshots, and reference documentation
Use Cases
- Cross-file refactoring and feature development in large codebases
- Developers who value version control and want AI edits to leave a Git record
- Developers who want to use local models (Ollama) to protect privacy
- Developers chasing maximum cost efficiency with low-cost models such as DeepSeek-V4.1-Flash
- Maintainers and contributors of open-source projects
Pros
- Completely free and open source: Apache-2.0 license, with no usage restrictions
- Best-in-class Git integration: automatic commits make version history obvious at a glance
- Maximum model freedom: not tied to any vendor, supporting both cloud and local models
- Strong on large projects: repo map technology understands the entire codebase
- Extremely cost-effective: paired with DeepSeek-V4.1-Flash, API costs can be driven very low
Pricing
Aider itself is completely free and open source (**Apache-2.0 license**). You supply your own LLM API key, and the actual cost depends on the model you choose—with DeepSeek-V4.1-Flash, off-peak input runs about ¥1 per million tokens and output about ¥4 per million tokens (¥2 and ¥8 respectively during peak hours), with additional discounts on cache hits. Local Ollama models can bring the cost to zero. Requires a Python 3.10–3.12 environment.
Summary
Aider is a benchmark tool for terminal AI programming—best-in-class Git integration, the freest model choice, and completely free and open source, with very low cost when paired with DeepSeek-V4.1-Flash. Note that its maintenance cadence slowed in 2026, with the latest release v0.86.2 dating to February 2026. The pure terminal experience also means a higher barrier to entry, making it a fit for developers with some experience. If you value version control and cost control, Aider remains worth trying.
Version History
- Aider v0.86.2 (current latest) (2026-02-12): Released v0.86.2 and has remained on that version, with the last code commit on the main branch dating to 2026-05-22; the update cadence slowed noticeably afterwards. Around 48.3k GitHub stars. The homepage's recommended model list still shows the Claude 3.7 Sonnet, DeepSeek R1, OpenAI o1, o3-mini, and GPT-4o generation, which lags currently in-service models—in practice you should check LiteLLM's model name mapping yourself.
- repo map v3 and Mature IDE Bridging (2025-12-10): repo map v3 rewrote the context selection strategy for large repositories, improving cross-file refactoring stability; VS Code / Cursor IDE bridging reached day-to-day usable status.