Plandex is a terminal-based AI coding agent designed for large projects and real world tasks.
Plandex addresses the challenge of using AI to handle complex, multi-step coding work that spans many files. It works by maintaining a cumulative diff review sandbox that keeps AI-generated changes separate from your project until you approve them, allowing you to review, roll back, and debug incrementally. The tool can work with up to 2 million tokens of context directly, or index larger directories using tree-sitter project maps. It supports full autonomy—loading files, planning changes, executing commands, and debugging automatically—while also offering fine-grained control for developers who prefer step-by-step review.
Plandex suits teams and individual developers tackling substantial features or applications where context size and project complexity have limited other AI tools. It works with models from Anthropic, OpenAI, Google, and open source providers, letting you combine different models within a single workflow. The tool is particularly valuable if you've encountered limitations with other AI assistants once your project grows beyond a certain size or changes become too intricate to handle in a single pass.
The project maintains active development across both its CLI and server components, with regular releases for each. Pull requests are welcomed, and the tool supports local self-hosted deployment as an alternative to cloud-based operation.
