The open-source, terminal-based AI coding agent that plans before it codes.
Plandex is an MIT-licensed AI coding agent that runs in your terminal and specializes in long, multi-file tasks โ refactors, migrations, feature implementation. Where most coding AIs jump straight to a diff, Plandex builds an explicit plan first, lets you review and revise it, then executes step-by-step with automatic sandboxing and version control. You bring your own API keys (OpenAI, Anthropic, Google, OpenRouter) so you only pay for what you use.
Who it's for: Backend/full-stack engineers, solo founders, and open-source maintainers who prefer the terminal, want to control costs via BYOK, and tackle tasks too large for a single inline diff.
Plandex drafts an explicit task plan before touching code. You edit, reject, or approve steps before execution โ no more surprise 30-file diffs that you have to undo.
Changes are applied to an isolated branch by default. See a diff, run tests, and only commit when you're satisfied. Bad states are one command to rewind.
Plug in OpenAI, Anthropic, Google, or OpenRouter API keys. No per-seat SaaS pricing โ you pay the model provider directly, which is dramatically cheaper for power users.
Single binary install, works over SSH, in Docker, in CI. Pairs naturally with tmux, vim, and any editor you already use โ no IDE lock-in.
Plandex fills the gap between "Copilot-style autocomplete" and "agentic IDE" โ it's the right tool when a task is too big for a single diff but you still want to stay in your terminal and your own editor. Pair it with a free completion tool and a $5 API credit, and you've got a serious Cursor alternative for under $10/mo.
The other big terminal-first AI pair programmer. Git-native, multi-file, runs offline.
The IDE-first AI code editor. Best for engineers who want completion + agent in one app.
Open-source AI code assistant for VSCode and JetBrains. BYOK, any model.
Open-source autonomous coding agent. Browser-based, runs full dev environments.