The terminal AI coding agent built for large, real-world tasks — a 2M-token effective context and a sandbox where you review every diff before it lands. A powerful Claude Code alternative.
Plandex is a terminal-based AI coding agent designed for big tasks — the kind that span many files and would blow past a normal context window. It manages an effective context up to around 2M tokens and keeps your changes in a protected sandbox, so nothing touches your project files until you approve it.
That review-first design is its signature: you can let the agent plan and generate extensive changes, then step through diffs and apply exactly what you want. You bring your own model and can self-host the whole thing, including the server and database.
It's licensed under MIT and AGPL (open-source server), with a focused, active community. For developers tackling large, multi-file changes who want an agent with serious context and careful, reviewable output, Plandex is a standout open alternative to Claude Code.
Yes — Plandex is open source (MIT and AGPL). You pay for the model you connect and, if self-hosting, your own server.
Plandex is open source, self-hostable and built around a huge context and a review-first sandbox for large tasks, while Claude Code is tied to Anthropic's platform.
It manages an effective context up to roughly 2 million tokens, making it suited to large, multi-file changes.
No — changes are held in a sandbox so you review each diff before applying, keeping your project safe.
Yes — you can self-host the full stack, including the server and PostgreSQL database, and bring your own model.
Terminal AI pair programmer that edits your codebase with any model.
Open-source AI coding agent for the terminal with 75+ LLM providers.
Autonomous AI dev agents that plan and ship code end-to-end.