Plandex is an open-source, terminal-based AI coding agent built for large projects and real-world development tasks. Rather than answering one prompt at a time, it plans multi-step work, executes it across dozens of files, and keeps every change in a review sandbox until you approve it. The project is developed in the open on GitHub, where it has earned more than 15,000 stars, with the most recent activity in 2025.
Plandex is designed for the kind of tasks that overflow a chat window: refactors, feature builds, and migrations that touch many files at once. It can load up to 2 million tokens of context directly and index project directories of 20 million tokens or more using tree-sitter project maps, so the agent works from a real map of your codebase instead of guessing. Changes accumulate as diffs in a sandbox, separate from your working tree, and commands run under your control, which makes it easy to review, roll back, and debug. It works with models from OpenAI, Anthropic, Google, and others, and you can run it against a hosted cloud server or fully self-hosted in local mode.
Plandex is free and open source under the MIT license. You can self-host the server locally at no cost beyond the model API usage you choose to pay for. A hosted Plandex Cloud option is also available for teams that prefer not to run their own server, which makes the overall model freemium.
Plandex suits professional developers and small teams tackling substantial changes in large codebases, especially anyone who wants an autonomous coding agent with a strict review step before changes land. If you like working in the terminal and want Claude Code-style autonomy with a diff sandbox and rollback, Plandex is one of the most capable open-source options available.