Goose

Block's open source local AI Agent framework can autonomously perform complete engineering tasks

Agent

https://github.com/block/goose

Visit Website

In-depth Report

Goose is a local AI Agent framework open sourced by Block Company (formerly Square) in early 2026, which can autonomously perform complete engineering tasks. Unlike traditional code completion tools, Goose can not only generate code, but also independently complete the complex development process from file reading, code writing, running tests to API calls, and can self-correct after errors. The project is written in Rust, has over 35,000 stars, and is licensed under the Apache-2.0 open source license. Goose provides both desktop application and CLI forms, supporting any LLM and MCP server integration.
Goose is developed by Block Company. Block is the parent company of the American payment company Square. It changed its name to Block in 2025 to reflect its business layout beyond payments. The company has always focused on developer tools, and Goose is its key product in the field of AI programming. The project is written in Rust language and focuses on performance and cross-platform compatibility. Goose was officially open sourced in early 2026 and quickly gained community attention. As of April 2026, the project has received more than 35,000 stars on GitHub, making it one of the fastest-growing open source AI programming tools. The project is actively maintained, with more than 4,000 submissions, and has an active community ecosystem.
Goose provides a variety of usage forms to meet the needs of different scenarios. The first is CLI mode, suitable for end users and developers who prefer the command line. The second is the desktop application mode, which provides a graphical interface to lower the threshold of use. The third is MCP server integration, which can be seamlessly connected with other development tools. In terms of core functionality, Goose is able to perform complete engineering tasks autonomously. It can read existing code, analyze project structure, write new code, run tests, and call external APIs. Each step can be automatically debugged and corrected after errors occur until the task is completed. This end-to-end automation capability is its core feature that distinguishes it from traditional code completion tools. Goose supports any LLM, including mainstream models such as OpenAI, Anthropic, and Google, as well as locally deployed models. Users can flexibly configure multiple models and select appropriate models based on task complexity to optimize performance and cost. The project deeply integrates the MCP protocol and can connect to many third-party service expansion capabilities. In terms of user experience, Goose design focuses on simplicity and ease of use. New users can complete configuration in five minutes through the quick start guide, and experienced developers can drive directly through natural language instructions. The desktop application provides a graphical configuration interface, lowering the technical threshold. Multi-model support and flexible configuration are highlights for advanced users.
Goose is an open source free project licensed under the Apache-2.0 license. This means that anyone can use, modify and distribute the project code for free. There is also no charge for commercial use, as long as you comply with the license requirements. While basic functionality is free, Block may be monetized through enterprise services and customized support. For example, it provides enterprise-level deployment solutions, dedicated technical support, advanced function plug-ins, etc. This model of open source core plus enterprise value-added is very common in the field of developer tools.
Judging from community feedback, Goose has received widespread praise. Users generally believe that its automation capabilities are strong and can significantly improve development efficiency. Many developers shared their experience of using Goose to complete the entire project, from the front-end page to the back-end API. Git workflow automation has been well received, and it is believed that it truly realizes the experience of "saying a requirement and the code is completed". The active open source community is another highlight. There are many issues and PR discussions on GitHub, and new features are updated frequently. The Discord community is active and the official response is timely. There are also many tutorials and discussions in the Chinese community, which are friendly to domestic users. Some users also put forward suggestions for improvements. For example, you want to add more model support, optimize local model compatibility, improve Windows compatibility, etc. Some users report that complex projects still have failure rates and require more manual intervention.
In the field of open source AI agents, Goose is one of the most popular projects. Competes with commercial products such as Claude Code and Codex CLI. Its open source nature and flexibility are its key differentiators. Industry analysts believe that open source AI Agent is an important direction for AI programming tools. The emergence of Goose has promoted the development of the field, and other projects are also rapidly iterating. Deep integration with the MCP protocol is considered a forward-looking technical decision, laying the foundation for building a richer tool ecosystem.
As an emerging technology, Goose faces some controversy. The first is security risk. AI agents executing code autonomously may bring security risks, especially accessing sensitive data or performing dangerous operations. Users are advised to use it in a controlled environment and pay attention to permission control. The second is dependency risk, relying solely on AI to generate code may affect development when the model is unavailable. Although the project supports local models, it has hardware requirements. Some users reported that complex tasks still failed and required patient debugging.
Goose is suitable for the following groups: developers who need to automate the development process, technology enthusiasts exploring AI programming tools, open source tool enthusiasts, and independent developers who do not require enterprise-level support. It is not suitable for enterprise users who are extremely sensitive to code security and users with limited hardware resources. For beginners, it is recommended to start with the desktop application and gradually explore the CLI and advanced features. Understand permission control before use and configure the AI model appropriately. Enterprise users should evaluate security needs and formulate usage specifications.
Goose is Block's open source local AI Agent framework, which provides capabilities from code completion to complete task automation. The 35,000-star open source recognition proves its technical value and community recognition. For developers looking for open source AI programming solutions, Goose is worth paying attention to. Its integration with the MCP protocol and flexible multi-model support provide the foundation for building personalized development workflows. Expect to see updates with more enterprise-grade functionality and security features in the future.

User Reviews

KAmey

—

Goose is really great and helped me complete the entire back-end project in three days.

InesJensen

—

35,000 stars is not a guarantee, it is indeed the benchmark for open source AI Agents.

realSaanaKilpela_dev

—

Multi-model configuration is very flexible, and the appropriate model can be selected according to the difficulty of the task.

DorothyJimenez

—

MCP integrates seamless connections and has strong scalability.

Henry.Turner520

—

The desktop application is friendly to newcomers, and you can get started immediately after being configured in five minutes.

KyleBell007

—

It is open source and free, and there is no cost for enterprise use. Thank you!

LIwil

—

More suitable for enterprises than Codex, with controllable security.

i2knq0

—

It has strong automation capabilities and can self-correct when errors occur, saving worry.

DSullivanJr1

—

The Discord community is active and official responses are quick.

Brandon68

—

The code generation quality is high and more stable than Claude Code.

黎强飞

—

Written in Rust, it performs very well and does not occupy resources.

Gloria_Lee520

—

The Chinese documentation is getting more and more complete, I’m touched.

Pamela_PowellK9

—

The local model supports good reviews and can be used even when the network is disconnected.

谭萍

—

Complex projects are still a bit difficult and require more debugging.

Alan.PhillipsQ

—

Workflow automation yyds, all done with one command.

ZGonzalez_2021

—

Access control must be done well, and safety comes first.

GEkra

—

Apache 2.0 license, no pressure for commercial use.

Sharon.Gonzales

—

Submissions on GitHub are active and updates are frequent.

DolphinDefiMartin

—

CLI is very convenient to use, and the terminal party is ecstatic.

ChristinaScott

—

The test autorun function is very useful.

Goose

In-depth Report

User Reviews

Similar Products