Skip to content
← Tools / Skills
ToolPlandex/10 trustMetrics-checked

Plandex

Terminal-based AI coding engine for large multi-file tasks, with a 2M-token effective context window.

Visit official site →

About

Plandex is an open-source, terminal-based AI coding engine built in Go, designed for large coding tasks that span dozens of files and many steps. It plans changes before implementing them, keeps AI-generated edits in a cumulative diff sandbox so you can review and selectively apply them, and builds fast tree-sitter project maps across 30+ languages so it stays reliable in big codebases. Model packs let you mix and match providers — Anthropic, OpenAI, Google, DeepSeek, Ollama and more — with context caching to cut cost and latency, and autonomy is configurable from full-auto down to fine-grained step-by-step control. It can also automatically debug terminal commands (builds, linters, tests) and browser apps when Chrome is installed.

The honest picture: Plandex pioneered ideas the whole CLI-agent category later absorbed, but its founder has acknowledged he couldn't compete head-on with Claude Code and moved on — so long-term momentum is uncertain even though the open-source core remains usable. The hosted Cloud tier runs about $45/month, and the terminal-first workflow has a steeper learning curve than IDE-native assistants.

What it can do

  • 2M-token effective context window
  • Cumulative diff review sandbox for safe edits
  • Tree-sitter project maps across 30+ languages
  • Multi-provider model packs (Anthropic, OpenAI, Google, Ollama)
  • Configurable autonomy levels
  • Automated debugging of builds, tests, and browser apps
  • Git integration with auto-commit option

Setup

Install via the one-line CLI installer from the Plandex docs; bring your own model provider API keys or use Plandex Cloud.

Prerequisites

API keys for your chosen model providers (or a Plandex Cloud subscription); a terminal and a Git repo.

Use cases

ai-codingrefactoringcode-generation

Works with

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 50; Trending score 72/100 from researcher assessment.

★ 15,183 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.