goose is an open-source, general-purpose AI agent built in Rust, hosted under the vendor-neutral Agentic AI Foundation (AAIF) at the Linux Foundation. Originally developed to execute local automation and development tasks on a user's machine, goose operates across software engineering workflows, data analysis, research, and system administration. In daily operations, developers interact with goose through a command-line interface, a native desktop application (macOS, Linux, Windows), or by embedding it via its API. goose supports 15+ LLM providers, including Anthropic, OpenAI, Google Gemini, OpenRouter, Azure, Amazon Bedrock, and local models via Ollama. It can also connect to existing consumer web subscriptions (such as Claude or ChatGPT) through the Agent Client Protocol (ACP), allowing users to bring their own keys or existing accounts. What sets goose apart from standard terminal coding agents is its deep architecture around open standards and automation primitives. It incorporates first-class Model Context Protocol (MCP) support with an ecosystem of over 70 extensions, an adversary review security architecture that monitors tool execution for unsafe actions, and portable YAML "Recipes" that allow teams to script, version, and execute repeatable agentic workflows in CI or local environments. goose is free and open-source software distributed under the Apache 2.0 license. Users pay only for their own underlying LLM inference providers.
Extensible MCP Ecosystem — Connects natively to 70+ Model Context Protocol servers for access to local files, databases, APIs, browsers, and developer tools. - Multi-Provider & ACP Support — Compatible with over 15 LLM backends (OpenAI, Anthropic, Ollama, Bedrock) and connects directly to IDEs and subscription agents via Agent Client Protocol. - Portable Workflow Recipes — Automate multi-step tasks using version-controlled YAML recipe files that specify instructions, parameters, extensions, and subrecipes for local execution or CI. - Parallel Subagents — Spawns isolated, independent subagents for code review, research, or heavy file processing without polluting the primary session context. - Built-in Security Controls — Includes sandbox execution modes, granular tool permission gates, prompt injection detection, and an adversary reviewer to monitor dangerous actions. - Rust-Powered Multi-Platform Binaries — Ships as a high-performance, native CLI, desktop app, or embeddable API across macOS, Linux, and Windows.
Use cases
Use Case: Automated CI/CD Scripting and Code Auditing Developers can encode linting, refactoring, and security verification pipelines into reusable YAML recipes. These recipes run headless via the goose CLI in CI environments or local pre-commit hooks, using subagents to review changes against project standards.
Use Case: Tool-Augmented Terminal Development Engineers working in the terminal can attach MCP extensions (e.g., GitHub, local Postgres, web search) directly to goose. The agent can inspect database schemas, write migrations, test them against local services, and open pull requests in a continuous loop.
Use Case: IDE Integration via ACP Server Developers can run goose as a background Agent Client Protocol (ACP) server connected to editors like Zed, JetBrains, or VS Code. This enables code assistance using local Ollama models or remote providers directly within the editor buffer.
Who it is for
goose is ideal for developers, DevOps engineers, and teams looking for an open-source, highly extensible agent harness with granular tool permissions and vendor independence. It may be unnecessary for users who only require simple inline autocomplete without external tool execution or custom MCP integrations.
Install guide
Installation Install the goose CLI on macOS, Linux, or Windows (WSL) using the official installation script: bash curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash Alternatively, install the desktop application by downloading the binary package directly from the goose releases.
First Run Initialize goose and configure your preferred model provider (e.g., OpenAI, Anthropic, Ollama, or OpenRouter): bash # Start goose configuration or launch the interactive CLI goose session Export your desired API key in your environment if prompted: bash export ANTHROPIC_API_KEY="your-api-key" # or export OPENAI_API_KEY="your-api-key"
Example Session Run an interactive coding task with specific context in your repository: bash goose session # Inside the goose prompt: > Analyze the src/ directory, write unit tests for the authentication handler in tests/auth_test.rs, and run cargo test to verify they pass.
OpenCode — OpenCode is an open-source AI coding agent developed by Anomaly that operates across terminal environments, desktop applications, and IDE extensions.…
Plandex — Plandex is an open-source terminal-based AI coding agent designed to handle complex, multi-file engineering tasks in large codebases. Created by…
SWE-agent — SWE-agent is an open-source autonomous software engineering agent harness built and maintained by researchers from Princeton University and Stanford University.…
OpenHands — OpenHands (formerly OpenDevin) is an open-source autonomous AI software development agent and orchestration platform. It provides an extensible runtime environment…
Weaver — Weaver is an open-source agent harness and orchestration framework maintained by sean35mm on GitHub. Built on top of the Bun…
OSOP Agent Rules — OSOP Agent Rules is an open-source orchestration harness and rule repository developed by Archie0125 (osop.ai) designed to standardize agent behavior,…