Plandex - AI Code Assistants Tool

Overview

Plandex is an open-source, terminal-first AI coding agent built to plan and execute large, multi-step development tasks across real-world projects. It emphasises robust context management (an effective 2M-token context with the default model pack), fast project mapping and syntax validation via tree-sitter, and a cumulative diff review sandbox that keeps AI-generated changes isolated until you approve them. Plandex supports mixing models from OpenAI, Anthropic, Google, Ollama and other providers so you can balance cost, speed, and capability for each role in the pipeline. ([github.com](https://github.com/plandex-ai/plandex)) Plandex is designed for workflows where a single change touches dozens of files or requires multi-step plans: it can generate and branch plans, stage diffs for manual review, run and automatically debug terminal commands (builds, tests, linters, deployments), and operate either fully autonomously or in a tightly controlled, stepwise mode. It offers an easy one-line CLI installer and a self-hosted Docker local-mode (server + CLI) for teams that need privacy or long-running local instances. Note that the hosted Plandex Cloud service was announced as winding down on October 3, 2025; the open-source project and self-hosted options continue to be maintained. ([github.com](https://github.com/plandex-ai/plandex))

GitHub Statistics

  • Stars: 14,869
  • Forks: 1,069
  • Contributors: 22
  • License: MIT
  • Primary Language: Go
  • Last Updated: 2025-10-03T21:49:54Z
  • Latest Release: cli/v2.2.1

The GitHub repository shows strong community interest and ongoing engineering activity: ~14.9k stars and ~1.1k forks indicate wide adoption and experimentation. The repo lists approximately 1,483 commits and a healthy contributor base (dozens of contributors across the project), with active issues and pull requests indicating ongoing maintenance and community feedback. Recent release notes emphasize expanded model-provider flexibility (v2.2.0 added multiple built-in providers and provider fallback logic), which signals active feature development and responsiveness to multi-provider workflows. The project remains MIT-licensed and frequently updated; the author also communicated a change in cloud hosting strategy in October 2025. ([github.com](https://github.com/plandex-ai/plandex))

Installation

Install via docker:

curl -sL https://plandex.ai/install.sh | bash
plandex  # or `pdx` — start the REPL in any project directory
export OPENROUTER_API_KEY=YOUR_KEY_HERE  # (or set other provider keys)
git clone https://github.com/plandex-ai/plandex.git && cd plandex/app && ./start_local.sh  # Docker local-mode server
plandex sign-in  # choose 'Local mode host' when prompted and confirm http://localhost:8099

Key Features

  • Effective 2M-token context window for very large projects and files.
  • Tree-sitter project maps for fast indexing and syntax validation across 30+ languages.
  • Cumulative diff review sandbox keeps AI edits isolated until approved.
  • Configurable autonomy: full-auto mode or stepwise, reviewer-driven execution.
  • Automated debugging of terminal commands (builds, tests, linters, deployments).
  • Multi-provider model support (OpenAI, Anthropic, Google, Ollama, OpenRouter, etc.).
  • Project-aware REPL chat for brainstorming before implementation.
  • Git-aware workflows: commit generation, branching, and versioned plans.

Community

Plandex has a sizable and active open-source community: around 14.9k GitHub stars and ~1.1k forks, with multiple contributors and active issues/PRs. The project maintains a public Discord (≈700 members reported on the site) and a documentation site with quickstarts and local-mode guides. The project author announced winding down the commercial cloud offering on October 3, 2025, while the open-source codebase and self-hosting options remain available. Community discussion, releases, and docs indicate ongoing maintenance and community-driven contributions. ([github.com](https://github.com/plandex-ai/plandex))

Last Refreshed: 2026-01-09

Key Information

  • Category: Code Assistants
  • Type: AI Code Assistants Tool