Aider is an open-source command-line AI pair programming tool that lets developers collaborate with frontier and local LLMs directly inside a terminal. It was created in 2023 by Paul Gauthier, a former Inktomi engineer and serial founder, and is distributed under the Apache 2.0 license from the Aider-AI/aider GitHub repository, where it has accumulated more than 40,000 stars.
At a technical level, Aider builds a repository map of the user's Git project, sends relevant context to the chosen LLM, applies suggested edits as patches, and automatically commits changes with sensible commit messages. It supports Claude 3.5 and 3.7 Sonnet, GPT-4o and o-series models, DeepSeek, Gemini 2.5, Cohere Command, local models via Ollama and LM Studio, and many others, with the ability to mix models for architect and editor roles. It also includes voice-to-code, image inputs, web-page ingestion, linter and test integration, and a benchmark harness that Aider uses to publish leaderboards of LLM coding performance.
Aider is maintained as an independent open-source project rather than a venture-backed startup, and total external funding is effectively zero. Paul Gauthier remains the primary maintainer, with a growing community of contributors driving features and language support. Usage and adoption stand in for traditional traction metrics: as of 2025-2026, the project reports millions of installs, with strong activity on GitHub, Discord, and developer forums.
From a workflow perspective, Aider's main strength is its uncompromising terminal-first design and tight Git integration. Developers can run it inside any project, ask for changes in natural language, watch Aider write code and tests, and immediately review the resulting commits. Because it is open source, teams can audit prompts, choose model providers, and self-host with local models for privacy.
Versus Claude Code, Gemini CLI, and Cursor, Aider's differentiators are its long track record, large user community, open-source license, and model-agnostic design. Its main trade-offs are a more spartan UX than IDE-style tools, dependence on developer-supplied API keys, and the need for users to manage model selection and cost themselves.