Jack Dorsey launches AI agent Goose, now you can do automatic coding
AI Agent: This agent is integrated with several large language models (LLMs) and can connect to various applications and tools through extensions. Currently, it can mainly handle engineering tasks, but the company is also exploring its non-engineering use cases.

The artificial intelligence (AI) agent "Codename Goose" was recently developed and released by Jack Dorsey's entity, Block. It is a parent entity to a mobile payments platform, Cash App, a point-of-sale platform, Square, and a streaming platform, Tidal. It recently developed an open-source AI agent for managing software development workloads. Several big language models (LLMs) have been incorporated with the agent, and it can integrate with many programs and tools through add-ons. It can manage engineering workloads at current times, but its non-engineering use cases are under development and research at present by the entity.
In a blog post, the Dorsey-led company announced "Codename Goose". It is an AI agent framework that can also be used to create new agents that can perform specific tasks. Since it is open-source, it can also be run locally, thereby reducing concerns related to data privacy. Due to its open-source framework, it provides versatile capabilities to developers. Goose can be powered by any LLM, whether it is an open-source model like DeepSeek-R1 or a property model like Gemini, Claude, or GPT.
In addition, this AI agent also supports a variety of extensions. These extensions allow Goose to connect to GitHub, Google Drive, JetBrains IDEs, and other tools and applications. The company said that some of these extensions have already been added to the agent's directory, while developers can also integrate and develop new extensions. Notably, all Goose extensions are based on the "Model Context Protocol (MCP)". Block also explained that Goose can be run locally as a desktop app or via a command-line interface (CLI) with the same configuration.
This AI agent is primarily used to handle coding and software development tasks and can carry out complex commands. Some of the major use cases of Goose include code migration (Ember to React, Ruby to Kotlin, etc.), conversion of code bases from field-based injection to constructor-based injection, performance benchmarking, creating Datadog monitors, adding or removing feature flags, etc.