Goose¶

What it is¶

Goose is an open-source AI agent framework developed by Block (formerly Square) that allows LLMs to interact with your local environment through a "toolkit" approach. It is designed to be a reliable, extensible assistant that can perform engineering tasks, manage infrastructure, and automate workflows.

What problem it solves¶

Goose addresses the need for a standardized, extensible way for agents to interact with system tools. By providing a clear interface for "tools" (executables, APIs, scripts), Goose allows developers to build agents that are grounded in real-world actions rather than just generating text.

Where it fits in the stack¶

Category: Automation & Orchestration / Agent Frameworks

Typical use cases¶

DevOps Automation: Automating deployments, checking logs, and managing Kubernetes clusters via CLI.
Engineering Assistance: Helping with code refactoring, running tests, and managing Git branches.
Tool Orchestration: Combining multiple disparate CLI tools into a single natural language workflow.

Strengths¶

Extensible Architecture: Easy to add new "skills" or tools to the agent's repertoire.
Local-First: Designed to run and act locally, ensuring speed and data privacy.
Block Ecosystem: Benefits from the engineering standards and patterns used at Block.
Modular: Separates the agent's reasoning (LLM) from the tools it uses.

Limitations¶

Newer Project: Community and ecosystem are still growing compared to older frameworks.
Complexity: Might require more setup than "zero-config" tools like Open Interpreter.

Getting started¶

Installation¶

Goose is typically installed as a CLI tool.

# Example (Check official repo for latest installation methods)
curl -fsSL https://github.com/block/goose/releases/latest/download/goose_install.sh | sh

Basic Usage¶

goose session

CLI examples¶

# Start a new interactive Goose session
goose session

# Run a specific task directly via the CLI
goose run "Scan the current directory for security vulnerabilities"

# List all available tools and skills in Goose
goose tools list

API examples¶

Goose is primarily a CLI-driven tool, but its toolset can be extended via the Model Context Protocol (MCP) or by writing custom Python toolkits.

# Custom tool definition for Goose (following MCP-like patterns)
def my_custom_tool(param: str) -> str:
    """
    A descriptive docstring that Goose uses to understand when to call this tool.
    """
    return f"Processed {param}"

# Tools are typically registered in a goose-config.yaml or via MCP

Sources / references¶

Contribution Metadata¶

Last reviewed: 2026-05-22
Confidence: high