Plandex
AI coding agent for multi-file tasks, with a Docker self-hosted server, MIT license, and support for Anthropic, OpenAI and Google models.
Plandex is an open source, terminal-based AI coding agent for developers working on changes that span many files. Its review sandbox holds generated edits apart from your project files so you can inspect the accumulated changes before applying them. You can run its server locally through Docker or host it on your own server. The code uses the MIT license.
Self-hosting the server doesn't make model calls local: Plandex uses your provider accounts and API keys, including OpenRouter, and sends model requests to those services. It supports models from Anthropic, OpenAI, Google and open source providers, with model packs that combine providers for different tasks. Windows support requires WSL. The hosted Plandex Cloud service isn't accepting new users.
For large codebases, Plandex uses tree-sitter project maps to identify relevant code and loads the context needed for each stage of a task. It checks syntax and can validate the logic of generated edits, with fallback approaches when an edit fails. Context caching for Anthropic, OpenAI and Google reduces repeated processing costs and latency.
You can let the agent plan changes, edit files, run commands and debug failures autonomously, or keep closer control through review. It can debug builds, linters and tests, plus browser applications when Chrome is installed. Project-aware chat supports questions about the code before implementation. Plans have their own version history and branches for comparing approaches or models; Git integration includes commit message generation and optional automatic commits.
