Goose
Open source, extensible AI agent that goes beyond code suggestions
Quick Take: Goose
Goose is the Swiss Army knife of AI agents. While other tools excel at specific tasks (Cursor for editing, Claude Code for terminal coding), Goose's extensible architecture lets it adapt to your entire workflow. The ability to connect AI intelligence to Jira, Slack, GitHub, databases, monitoring tools, and any custom service through a simple plugin system is uniquely powerful. Being free, open source, and model-agnostic makes it accessible to everyone. If you want an AI agent that fits your workflow rather than forcing you into a new one, Goose is the most flexible option available.
What is Goose?
Goose is an open-source, extensible AI agent now stewarded by the Agentic AI Foundation (AAIF) at the Linux Foundation. Originally created by Block (formerly Square), Goose takes a unique approach to AI-assisted development: rather than being a code editor or chat interface, it's a general-purpose AI agent that can be extended with extensions to interact with any tool, service, or workflow you use. Think of it as a universal AI assistant that you can teach new skills. Out of the box, Goose can read and write files, execute shell commands, browse the web, manage git repositories, and interact with GitHub. But its real power comes from its extension system—you can add extensions via the Model Context Protocol (MCP) that let Goose interact with databases, browsers, APIs, cloud services, and any tool you can imagine. Built for developers who want an AI agent that adapts to their specific workflow rather than forcing them into a predefined IDE experience, Goose runs as a desktop app, CLI tool, or API. The desktop app provides a clean chat interface while behind the scenes Goose executes code, manages files, and orchestrates tools autonomously. With 38,000+ GitHub stars and 400+ contributors, Goose supports multiple LLM backends including Claude, GPT-4, Gemini, and local models via Ollama, giving you flexibility in model choice and complete data privacy.
Install with Homebrew
brew install --cask block-gooseKey Features
Extensible MCP Extension System
Goose's defining feature is its deep integration with the Model Context Protocol (MCP). With 70+ documented extensions available, you can connect Goose to databases, browsers, GitHub, Google Drive, APIs, cloud services, and more. Extensions can even render interactive UIs directly inside Goose Desktop—buttons, forms, and visualizations that create agent-powered tools. You can also build custom extensions to connect Goose to any internal tool your team uses.
Autonomous Task Execution
Goose works as an autonomous agent, not just a chatbot. Describe a task like 'review the open PRs on this repo and summarize which ones are ready to merge' and Goose will navigate GitHub, read the PRs, analyze the code changes, check CI status, and produce a structured summary—all without further input from you. Goose can execute shell commands, modify files, and orchestrate complex multi-step workflows.
Multi-Model & Subscription Support
Goose supports 15+ LLM providers including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock. Through the Agent Client Protocol (ACP), you can even use your existing Claude, ChatGPT, or Gemini subscriptions. This flexibility lets you choose the best model for your use case—or use local models via Ollama for complete data privacy when working with sensitive code.
Desktop, CLI, and API
Goose is available as a native desktop app for macOS, Linux, and Windows; a full CLI for terminal workflows; and an API to embed it anywhere. Built in Rust for performance and portability, all three interfaces share configurations and extensions, letting you seamlessly switch between visual and command-line workflows.
Recipes & Subagents
Capture complex workflows as portable YAML configs called Recipes—share them with your team, run them in CI, and include parameters and sub-recipes. Spawn independent subagents to handle tasks in parallel—code review, research, file processing—keeping the main conversation clean while dramatically speeding up multi-part tasks.
Security & Permissions
Goose includes multiple layers of security: prompt injection detection, tool permission controls, sandbox mode, and an adversary reviewer that watches for unsafe actions. Goose shows you exactly what it's doing at each step—which commands it's running and what files it's modifying—giving you full transparency and control.
Who Should Use Goose?
1Platform Engineer
A platform engineer needs to check the health of their microservices, review recent deployments, and identify any services with elevated error rates. They ask Goose, which uses its Kubernetes toolkit to query the cluster, its monitoring toolkit to check Datadog dashboards, and its Slack toolkit to check #incidents. Goose produces a comprehensive status report in 60 seconds—work that normally involves checking three different dashboards.
2Full-Stack Developer
A developer wants to set up a new feature branch, create a Jira ticket, scaffold the API endpoint and React component, and open a draft PR. They describe the feature to Goose in one message. Goose creates the Jira ticket (via the Jira toolkit), creates the git branch, writes the boilerplate code, commits it, and opens a draft PR on GitHub—orchestrating five different tools from a single conversation.
3Open Source Maintainer
An open source maintainer uses Goose to triage issues every morning. Goose reads new GitHub issues, categorizes them (bug, feature request, question), checks if they're duplicates of existing issues, and drafts initial responses. The maintainer reviews Goose's work and approves the responses—turning a 45-minute daily ritual into a 10-minute review session.
Install Goose on Mac
Goose is available through Homebrew as an open-source desktop application and CLI tool. You'll need an API key for at least one LLM provider (OpenAI, Anthropic, Google) or a local model via Ollama.
Install Homebrew
If you don't have Homebrew, open Terminal and run: `/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"`
Install Goose Desktop
Run: `brew install --cask block-goose`
(Optional) Install Goose CLI
For terminal workflows, also run: `brew install block-goose-cli`
Configure LLM Provider
Launch Goose and configure your preferred LLM provider. Add your API key for Claude, GPT-4, Gemini—or connect to a local Ollama instance for privacy-first usage. Use `goose configure` in CLI.
Configuration Tips
Enable the Toolkits You Need
Goose ships with several built-in toolkits (file system, shell, git, GitHub). Enable only the toolkits you need for security and focus. Add specialized toolkits like Jira, Slack, or database connectors as needed.
Try Local Models for Sensitive Code
If you work with proprietary or sensitive code, configure Goose to use a local model via Ollama. This keeps all code processing on your machine while still benefiting from AI assistance.
Write Custom Toolkits
Goose's toolkit API is straightforward Python. If your team uses internal tools that Goose doesn't support out of the box, write a custom toolkit to connect them. This turns Goose into a universal interface for your entire development ecosystem.
Alternatives to Goose
Goose occupies a unique position as an extensible AI agent, competing with both coding assistants and general-purpose AI tools.
Claude Code
Claude Code is a terminal-based coding agent focused on code reading, writing, and execution. Goose is broader—it can do coding tasks but also interact with Jira, Slack, monitoring tools, and any custom integration. Claude Code is deeper for pure coding; Goose is wider for workflow orchestration.
ChatGPT
ChatGPT is a conversational AI without direct tool access (beyond plugins). Goose can autonomously execute commands, modify files, and interact with external services. ChatGPT is better for knowledge questions; Goose is better for tasks that require action.
Raycast AI
Raycast AI provides quick AI access through the Raycast launcher. It's fast for one-off questions but lacks Goose's ability to execute multi-step autonomous tasks across multiple tools and services.
Pricing
Goose is **completely free and open source** (Apache 2.0 license). You provide your own LLM API keys, so costs depend on your model usage. Using local models via Ollama means zero ongoing cost beyond your hardware. This makes Goose the most cost-effective AI agent platform available.
Pros
- ✓Free and open source with no subscription fees
- ✓Extensible toolkit system connects to any tool or service
- ✓Multi-model support including local models for privacy
- ✓MCP support for standardized integrations
- ✓Autonomous agent capabilities beyond simple chat
Cons
- ✗Requires your own LLM API keys (cost management is your responsibility)
- ✗Custom toolkit development requires Python knowledge
- ✗Less polished UX compared to commercial products like Cursor
- ✗Smaller community than major AI tools
Community & Support
Goose is an open-source project from Block with an active GitHub repository, community Discord, and growing contributor base. The project welcomes contributions—especially new toolkits that extend Goose's capabilities. Documentation covers toolkit development, configuration, and model setup.
Frequently Asked Questions about Goose
Our Verdict
Goose is the Swiss Army knife of AI agents. While other tools excel at specific tasks (Cursor for editing, Claude Code for terminal coding), Goose's extensible architecture lets it adapt to your entire workflow. The ability to connect AI intelligence to Jira, Slack, GitHub, databases, monitoring tools, and any custom service through a simple plugin system is uniquely powerful. Being free, open source, and model-agnostic makes it accessible to everyone. If you want an AI agent that fits your workflow rather than forcing you into a new one, Goose is the most flexible option available.
About the Author
Expert Tips for Goose
Start with the built-in toolkits (file, shell, git, GitHub) before adding custom ones. Once you're comfortable with Goose's agent behavior, write a toolkit for your team's most time-consuming workflow—the ROI is immediate.
Related Technologies & Concepts
Related Topics
AI Coding Agents
Autonomous AI agents that can execute code, manage files, and interact with development tools.
Sources & References
Key Verified Facts
- The official GitHub repository confirming that Goose is an open-source AI agent developed by Block (formerly Square) designed to execute shell commands and manage local workflows.[cite-1]
- Official release notes documenting Goose's native support for macOS and its standalone desktop application capabilities.[cite-2]
- Official documentation defining Goose as a general-purpose AI agent that differs from standard chat interfaces by operating as an extensible, tool-agnostic assistant.[cite-3]
- Details how Goose can be taught new skills and extended with custom plugins to interact with external services, APIs, and specific developer workflows.[cite-4]
- Outlines Goose's out-of-the-box features, specifically noting its ability to read and write files, execute shell commands, browse the web, and manage git repositories.[cite-5]
- 1block/goose: An open-source, extensible AI agent
Accessed Mar 1, 2026
"The official GitHub repository confirming that Goose is an open-source AI agent developed by Block (formerly Square) designed to execute shell commands and manage local workflows."
- 2Releases · block/goose
Accessed Mar 1, 2026
"Official release notes documenting Goose's native support for macOS and its standalone desktop application capabilities."
- 3What is Goose? | Goose Documentation
Accessed Mar 1, 2026
"Official documentation defining Goose as a general-purpose AI agent that differs from standard chat interfaces by operating as an extensible, tool-agnostic assistant."
- 4Plugins and Extensibility | Goose Documentation
Accessed Mar 1, 2026
"Details how Goose can be taught new skills and extended with custom plugins to interact with external services, APIs, and specific developer workflows."
- 5Core Capabilities | Goose Documentation
Accessed Mar 1, 2026
"Outlines Goose's out-of-the-box features, specifically noting its ability to read and write files, execute shell commands, browse the web, and manage git repositories."
- 6Block open-sources Goose, an AI agent for developers
Accessed Mar 1, 2026
"TechCrunch report on Block releasing Goose as an open-source project, highlighting its unique approach to AI-assisted development outside of traditional code editors."
- 7Block Introduces Goose: An Open-Source Extensible AI Agent
Accessed Mar 1, 2026
"Technical analysis of Goose's architecture, discussing its extensibility and how it functions as a universal AI assistant rather than being tied to a specific IDE."
- 8Block open-sources Goose, an extensible AI agent | Hacker News
Accessed Mar 1, 2026
"Community discussion and verification of Goose's practical utility in seamlessly managing git repositories and executing shell commands on macOS."
- 9Evaluating Goose: Benchmarking Block's New AI Agent
Accessed Mar 1, 2026
"Block's developer blog detailing performance benchmarks of Goose's file parsing, tool execution speeds, and context management on Apple Silicon macOS devices."