Plandex

Open-source terminal AI coding agent built for large, multi-file projects

Hand-reviewed by the Toolglade team·Updated Aug 2026

coding#ai-coding-agent#open-source#terminal#cli

Free plan Claimed Self-hosted

About Plandex

Plandex is an open-source terminal AI coding agent optimized for large multi-file tasks, using a diff review sandbox and huge context via project maps, and running on models from multiple providers with a self-hosted, free workflow.

Plandex targets the workflow gap between an IDE assistant like Cursor and a general-purpose terminal agent: it is purpose-built for planning, reviewing, and applying changes across dozens of files. It can handle up to roughly 2M tokens of context directly (about 100k per file) and index large directories with tree-sitter project maps that reach 20M tokens or more, making it well-suited to sizeable codebases. A core design principle is reviewability. Plandex keeps AI-generated edits in a cumulative diff sandbox that stays separate from your working files until you decide to apply them, and it controls command execution so changes can be rolled back and debugged safely. It combines strong models from Anthropic, OpenAI, Google, and open-source providers within a robust terminal-based loop. The project is MIT-licensed with 14,000+ GitHub stars and an active community. It previously offered a hosted Plandex Cloud service, which is winding down in 2026 in favor of self-hosted and local modes, keeping the tool free to run for developers who bring their own model keys.

TL;DR

Plandex is a free, MIT-licensed terminal AI coding agent optimized for large multi-file tasks, emphasizing reviewable diffs and huge context, and running on models from multiple providers.

Company overview

Plandex is an open-source project (Plandex AI) that has grown to 14,000+ GitHub stars by focusing on a niche between IDE assistants and general terminal agents. Its design centers on planning, reviewing, and applying AI changes across large codebases.

The project maintains a free, self-hosted core and has chosen to wind down its hosted Cloud offering in 2026, reinforcing a community-driven, bring-your-own-key model.

Product features

Plandex runs in the terminal and builds a plan for a task, generating edits into a cumulative diff sandbox that stays separate from your files until approved. It supports very large context via tree-sitter project maps and controlled command execution with rollback.

It is model-agnostic, combining Anthropic, OpenAI, Google, and open-source models, letting developers optimize for cost and capability while keeping changes fully reviewable.

Target market

Plandex targets terminal-first developers and teams working on large, real-world codebases who want an open-source, self-hosted agent with strong reviewability and large context handling.

Buyer personas

End users

Individual developers and engineers working in the terminal on large multi-file tasks.

Buyers

Small teams and self-hosting engineers choosing open-source agent tooling.

Key influencers

Open-source contributors and terminal-workflow advocates.

Ideal customer profile

Experienced developers on large codebases who value self-hosting, reviewable diffs, and model flexibility.

Funding & performance

Plandex is a community-driven open-source project; no major institutional funding is publicly emphasized. Verify any funding details with the maintainers.

Pros & cons

Pros

Purpose-built for large, multi-file tasks
Cumulative diff sandbox keeps changes reviewable
Very large context via tree-sitter project maps
Model-agnostic across major and open providers
MIT-licensed and free to self-host
Command execution with rollback and debugging

Cons

Terminal-only — no GUI or IDE integration
Hosted cloud service is being wound down in 2026
Requires bringing your own model API keys
Smaller ecosystem than IDE-based competitors
Learning curve for the plan/review workflow

Pricing plans

Open Source (Self-Hosted / Local)

MIT-licensed CLI
Cumulative diff sandbox
Tree-sitter project maps
BYO model API keys
Command execution with rollback

Key features

Self-hosted

Multi-language

Integrations

Anthropic, OpenAI, Google Gemini, OpenRouter, Git

Input types

text

Output types

text

Best For

large multi-file coding tasks, terminal-centric workflows, reviewable AI changes

Compare key features

View all alternatives →

Feature

Plandex

Refact.ai

LiteLLM

Pricing

Freemium

Free plan

Yes

Free trial

Yes

API

Yes

Self-hosted

Yes

Team support

Yes

Frequently asked questions

Is Plandex free?+

Yes. Plandex is open-source under the MIT license and free to self-host or run locally. You pay only for the model API usage from your chosen providers.

How is Plandex different from Cursor or Claude Code?+

Plandex is terminal-based and purpose-built for large multi-file tasks, sitting between an IDE assistant like Cursor and a general-purpose terminal agent, with a strong focus on planning and reviewable diffs.

How much context can Plandex handle?+

It supports up to roughly 2M tokens of direct context (about 100k per file) and can index large directories with tree-sitter project maps reaching 20M tokens or more.

Which models does Plandex support?+

It combines models from Anthropic, OpenAI, Google, and open-source providers, configurable through providers such as OpenRouter.

Is there still a hosted cloud version?+

Plandex previously offered a hosted Cloud service that is winding down in 2026 in favor of self-hosted and local modes.