Block's new open-source AI agent 'goose' lets you change direction mid-air

Whether for developer needs or mundane tasks, the artificial intelligence (AI) tide appears to be turning in favor of open-source solutions. 

On Tuesday, Block — the Jack Dorsey-founded company behind Square, Cash App, Tidal, and Afterpay — released a total rewrite of codename goose, its open-source AI developer agent previously available in beta. Released under the Apache License 2.0 (ASL2), codename goose offers “interoperability between user interfaces, language models, and systems,” as the release explains, and connects to any system via Anthropic’s Model Context Protocol (MCP), which lets agents access data stores, developer spaces, and business applications. Users can specify any large language model (LLM) they want to use with goose. 

Also: The best open-source AI models: All your free-to-use options explained

Available for any individual or commercial purpose, the framework lets users connect new tools midway during a session (as opposed to delineating them at the beginning, as is necessary with closed-source tools). 

“You can be in the middle of a conversation, realize that you need to pull in something from your Google Docs or connect it to Slack or something, and you can just, say, click through, enable a new extension, and then goose suddenly has these new capabilities it can go run with,” lead engineer Brad Axen explained to ZDNET in an interview.

Similar to developer-first platforms like Blitzy, goose’s primary use is as an engineering agent that searches and writes code in real-time while “executing tasks autonomously — reading and writing files, running code and tests, refining outputs, installing dependencies, and handling additional actions as needed,” the release claims.  

To make it embeddable, the team switched the agent’s language from Python to Rust and built an accompanying native application for which goose itself produced 70% of the code. Axen noted that codename goose “saves 20% of [developers’] time, or they’ll find tasks where it can do a day or two worth of work all by itself, unsupervised.”

Also: I tried Perplexity’s assistant, and only one thing stops it from being my default phone AI

That said, Block hopes goose will eventually be used for other pursuits, including creative ones like music generation. “We really built a lot of the fundamental use features as kind of like a general purpose capability,” explains Jackie Brosamer, head of AI at Block. “A big part of the way we’re able to expand beyond the developer use cases is our open source philosophy.” 

Last week, Block launched an Open Source Program Office to cement its strategic commitment to open-source resources, expertise, and more. 

Brosamer notes that integrating with Anthropic’s MCP lets users connect with daily apps like email, Google Calendar, and Drive. “We’re starting to see people use codename goose to do things like simplify your day and try to get summaries of everything you read next last week,” she adds. Within Block’s suite of financial tools, Brosamer envisions a customer using goose to have Cash App automatically split bills among their roommates, run the books on their business each month, or even order dinner through Square. 

Also: The future of sales? These AI agents never sleep

“We definitely see goose opening up a whole range of capabilities to less technical users,” Brosamer notes. “We’re already seeing designers able to prototype their own front-ends without having to bring in an engineer.” 

Beyond that, Block talks about goose joining the growing list of day-to-day AI assistants. “I could see a place where, as we get smaller models that can do the agent loop effectively, we start to move those on device,” Axen says. “It’s pretty easy for us to take the agent path of goose, accompany it with an on-device model, and put it onto an Android phone or something like that.” 

Codename goose comes shortly after the release of DeepSeek’s R1 — an open-source model that challenges OpenAI’s o1 for a fraction of the cost — as well as AI assistants from both Perplexity and OpenAI. It’s well-timed, as the agent aims to fill both commercial and personal use needs, all while leveraging the endless possibilities created by open-source. 

Also: Operator isn’t worth its $200-per-month ChatGPT Pro subscription yet – here’s why

“If someone has a good idea, they don’t have to wait on us to build it,” Axen says. 

“Anyone can contribute new capabilities to the agent itself,” Brosamer adds. 

You can access codename goose along with all documentation here.