Goose
Open-source local AI agent for engineering tasks
Goose is profiled here as a Testing tool for engineering teams. Read about features, pricing, and how it compares to related options in the tools directory.
Description
Goose is an open-source AI agent from Block, the company behind Square and Cash App, released in early 2025. It runs on a developer's machine and executes real work end to end: writing code, running commands, debugging failures, and orchestrating tasks through MCP extensions. Goose connects to any LLM the team chooses, so it works with frontier APIs or self-hosted models without lock-in. Block released it under an open license so teams can extend it freely. Because the agent runs against local models, it supports offline and privacy-sensitive work where sending code to a hosted API is not an option.
Key Capabilities:
Local autonomous agent that edits files and runs shell commands
Model-agnostic backend across Anthropic, OpenAI, and self-hosted LLMs
MCP extensions for connecting external tools and data sources
Desktop application and command-line interface
Recipes that package repeatable agent workflows for sharing
Apache 2.0 license with no required cloud account
Alternative tools
- Storybook
Workshop for building and documenting UI components in isolation
- Zencoder
Repository-aware coding and unit-testing agents in your IDE
- Keploy
Generate API tests and mocks from real traffic
- Cypress
Developer-friendly end-to-end testing that runs in the browser
- Playwright
Cross-browser end-to-end testing and browser automation
- GX (Great Expectations)
Declarative data quality testing for pipelines