Goose (by Block)

Open-source, model-agnostic AI coding agent for CLI and desktop

Hand-reviewed by the Toolglade team·Updated Aug 2026

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

Free plan Claimed API Self-hosted

Toolglade’s take

Goose is a genuinely compelling free alternative to paid coding agents: fully open source, local-first, model-agnostic, and with excellent MCP support baked in. Block's backing and the move to the Linux Foundation give it real staying power and governance. Because it is BYO-model, you still pay for whatever LLM API you point it at, and setup is more hands-on than a polished commercial IDE. Non-technical users and teams wanting turnkey support may prefer a hosted product.

About Goose (by Block)

Goose is Block's open-source, Rust-built, model-agnostic AI coding agent that runs locally via CLI or desktop, autonomously executes dev tasks, and has first-class MCP support.

Goose is a CLI-first and desktop AI coding agent open-sourced by Block (the company behind Square, Cash App, and Tidal). It autonomously executes software development tasks end to end, running shell commands, debugging errors, orchestrating multi-step workflows, and even building complete projects from scratch. Written in Rust, it runs entirely on the developer's local machine as a desktop app, CLI, or API. Goose is model-agnostic, connecting to Claude, GPT, Gemini, local models via Ollama, or any OpenAI-compatible endpoint, so teams avoid vendor lock-in. It has first-class Model Context Protocol support: it can both consume MCP servers as tools and expose itself as an MCP server for other agents, and Block ships official MCP servers for GitHub, PostgreSQL, Slack, and Jira. The project is fully open source under Apache 2.0 with no paid tier. Donated to the Agentic AI Foundation under the Linux Foundation, Goose has grown to tens of thousands of GitHub stars and hundreds of contributors. It suits developers who want a free, private, extensible coding agent that runs locally and plugs into the MCP ecosystem, as an alternative to hosted proprietary agents.

TL;DR

Goose is Block's free, open-source, Rust-built AI coding agent that runs locally, works with any LLM, and has first-class MCP support, governed under the Linux Foundation.

Company overview

Goose was created by Block, the fintech company behind Square, Cash App, and Tidal, as an internal coding agent before being open-sourced in 2026. It was subsequently donated to the Agentic AI Foundation under the Linux Foundation for neutral governance.

The project has grown rapidly to tens of thousands of GitHub stars and hundreds of contributors, with active daily development. As an open-source project it has no commercial pricing of its own.

Product features

Goose autonomously executes development tasks, running shell commands, editing files, debugging, and orchestrating multi-step workflows. Built in Rust, it runs locally as a CLI, desktop app, or API on macOS, Linux, and Windows, and is model-agnostic across cloud and local LLMs.

Its standout capability is first-class MCP support, consuming MCP servers as tools and exposing itself as one, with official servers for GitHub, PostgreSQL, Slack, and Jira. This makes it highly extensible within the growing agent ecosystem.

Target market

Goose targets software developers and engineering teams who want a free, private, extensible coding agent that runs locally, avoids model lock-in, and integrates with MCP tooling.

Buyer personas

End users

Software developers using CLI and desktop coding agents.

Buyers

Engineering leaders adopting open-source agent tooling.

Key influencers

Open-source advocates, MCP ecosystem builders, and Rust developers.

Ideal customer profile

Developers and teams wanting a free, local-first, model-agnostic coding agent with strong MCP support and willingness to self-manage setup and model costs.

Funding & performance

Open-source project backed by Block and the Linux Foundation; no separate venture funding (verify with the project).

Pros & cons

Pros

Completely free and open source under Apache 2.0
Model-agnostic with no vendor lock-in
Runs locally for privacy and control
First-class MCP support both ways
Backed by Block and the Linux Foundation
Available as CLI, desktop app, and API

Cons

You still pay for the underlying LLM API
Setup is more hands-on than commercial IDEs
No built-in team collaboration features
Autonomous execution needs careful supervision
Less polished UX than paid competitors

Key features

API

Self-hosted

Multi-language

Integrations

GitHub, PostgreSQL, Slack, Jira, Ollama, MCP servers

Input types

text

Output types

text

Best For

Autonomous coding tasks, Local-first development, MCP-based workflows

Compare key features

View all alternatives →

Feature

Goose (by Block)

Refact.ai

Cline

Pricing

Free

Freemium

Free

Free plan

Yes

Free trial

Yes

API

Yes

Self-hosted

Yes

Team support

Yes

Frequently asked questions

Is Goose really free?+

Yes. Goose is fully open source under Apache 2.0 with no paid tier. You only pay for whatever LLM API or local compute you choose to use with it.

Which models does Goose work with?+

Goose is model-agnostic and works with Claude, GPT, Gemini, local models via Ollama, or any OpenAI-compatible endpoint.

Does Goose support MCP?+

Yes. Goose has first-class Model Context Protocol support and can both consume MCP servers as tools and expose itself as an MCP server. Block ships official servers for GitHub, PostgreSQL, Slack, and Jira.

Who maintains Goose?+

Goose was created and open-sourced by Block and donated to the Agentic AI Foundation under the Linux Foundation, with hundreds of contributors.

Can Goose run locally?+

Yes. Goose runs entirely on your local machine as a desktop app, CLI, or API, and can use local models for full privacy.