Dorsey launches Goose agent

Jack Dorsey’s Block put Goose into the open, turning an internal coding agent into a free project developers can run on their own machines. (block.xyz) Block announced Goose on January 28, 2025 as an Apache 2.0 open-source framework for AI agents, with its first use aimed at software engineering. The software can read and write files, run code and tests, install dependencies, and take follow-up actions inside a development environment. (block.xyz) The basic idea is simple: a large language model predicts text, and an agent adds tools so that text can trigger actions. Goose connects models to outside systems through Model Context Protocol, a shared interface for tools and data that Block said it developed closely with Anthropic. (block.xyz) Goose is not limited to one model vendor or one interface. Its current documentation says it runs as a desktop app, command line tool, and application programming interface, works with more than 15 providers, and supports more than 70 Model Context Protocol extensions. (goose-docs.ai) That puts Goose in a different lane from cloud coding assistants that sell access as a hosted service. Block’s pitch is that developers can keep the agent on-machine, connect it to the models and tools they already use, and modify the code because the project is open source. (goose-docs.ai) The project also comes with signs of real use beyond a launch post. Block’s AI page says Goose grew out of the company’s own work, and The New Stack reported in October 2025 that Block had rolled AI agents out to its 12,000-person workforce in eight weeks. (block.xyz) (thenewstack.io) Community interest is visible in the repository numbers. The GitHub project listed about 40,600 stars, roughly 4,000 forks, and more than 4,100 commits when checked on April 13, 2026. (github.com) Goose’s ownership also changed this month. On April 7, 2026, the project said Block had donated it to the Agentic AI Foundation at the Linux Foundation, and the code moved to the `aaif-goose` GitHub organization. (goose-docs.ai) The result is that Goose now sits as both a product of Dorsey’s company and a foundation-backed open-source project. For developers weighing whether to rent an assistant in the cloud or run one locally, that makes Goose easier to test and harder to ignore. (goose-docs.ai)