Is Goose free to use?

Yes, Goose is free and open source under the Apache 2.0 license. Users must provide their own model API keys or run local models via Ollama, with provider costs passed through.

How does Goose integrate with external tools?

It uses the Model Context Protocol (MCP) to attach capabilities from databases, browsers, filesystems, and cloud APIs. There are 70+ ready-made extensions available, and users can write their own.

Can I share or schedule my workflows in Goose?

Yes, the 'Recipes' system allows users to capture a working session as a portable YAML file. This makes workflows reproducible, shareable, and schedulable.

Who is Goose designed for?

It is best for developers and small technical teams who want a fully controllable, local-first agent. It is not intended for non-technical users seeking a plug-and-play chatbot or teams requiring managed SaaS collaboration features.

Goose

Open-source, local-first AI agent for code, workflows, and everything in between.

Free· Free and open source under Apache 2.0. You bring your own model API keys (or run local models via Ollama); provider costs are pass-through.AgentsProvider-agnostic — Claude 3.5/4, GPT-4o/4.1, Gemini 1.5/2, Llama 3.x via Ollama, plus Azure OpenAI, Bedrock, OpenRouter and others

Visit website →

In short

Goose is a provider-agnostic AI agent that runs locally as a desktop app, CLI, or API. It is best for developers who need a controllable, open-source tool for multi-step coding tasks and data analysis using MCP extensions.

Best for

Developers and small technical teams who want a fully controllable, local-first AI agent that can drive real tools via MCP and swap freely between frontier and local models.

Skip if

Non-technical users who want a plug-and-play chatbot, or teams that need managed SaaS collaboration, SSO, and billing rolled into the agent itself.

Goose is an open-source AI agent that runs on your own machine as a desktop app, CLI, or programmable API. It was originally built at Block and has since moved to the Agentic AI Foundation at the Linux Foundation for vendor-neutral governance, and it has grown into one of the larger community-driven agent projects (45k+ GitHub stars, 500+ contributors). Where hosted assistants lock you into a single model and toolset, Goose is deliberately provider-agnostic: it talks to 15+ LLM backends including Anthropic, OpenAI, Google, Azure OpenAI, AWS Bedrock, OpenRouter, and local models through Ollama, and it hosts its capabilities through the Model Context Protocol so any MCP server (databases, browsers, filesystems, cloud APIs, internal tools) can be attached as an extension. There are 70+ ready-made extensions and you can write your own. Typical workflows include multi-step coding tasks (planning, edits across a repo, running tests, opening PRs), research assignments that stitch together web fetch, scraping, and note-taking, data-analysis loops against local files or warehouses, and internal automations like triage or report generation. The 'Recipes' system lets you capture a working session as a portable YAML file so a workflow can be re-run, shared, or scheduled. Subagents let a parent agent spawn parallel workers for fan-out tasks, and interactive UI blocks let a recipe surface a small form or chart instead of a wall of text. Security features include prompt-injection detection on tool output and an adversary-review mode that critiques the agent's own plan before it executes destructive actions. It fits developers and small technical teams who want a Cursor/Claude-Code-style agent they fully control, can point at private data without shipping it to a SaaS, and can swap between frontier and local models depending on cost and sensitivity.

Editor's take

Goose is the most credible open-source answer to Cursor/Claude-Code-style agents right now: model-agnostic, MCP-native, and genuinely local. It rewards operators who like YAML and CLIs, and the safety posture is more thoughtful than most self-hosted agents. If you want a hosted, one-click experience, look elsewhere — but if you want an agent you actually own, this is the one to try first.

— The AI Tool Bible editorial team

Pros

✅ Truly model-agnostic — swap between Claude, GPT, Gemini, Bedrock, or a local Ollama model without changing your workflow
✅ First-class MCP support with 70+ prebuilt extensions and easy custom ones, so it plugs into real tools instead of being a chat box
✅ Runs locally as desktop app or CLI, so code, files, and prompts never leave your machine unless you route them to a hosted model
✅ Recipes (YAML) make agent runs reproducible, shareable, and schedulable rather than one-off chats
✅ Subagents enable parallel fan-out for search, refactor-across-repos, or batch analysis tasks
✅ Apache 2.0 open source with active governance under the Linux Foundation, so no vendor lock-in or surprise pricing
✅ Built-in prompt-injection detection and adversary review reduce blast radius when agents are given shell or write access

Cons

⚠️ You bring your own LLM keys and pay the provider directly — costs on frontier models can climb quickly for long agent loops
⚠️ Setup is heavier than a hosted chatbot: install app, configure provider, install and permission MCP extensions
⚠️ Local model quality (via Ollama) still lags frontier models for complex coding or tool-use chains
⚠️ Fewer polished collaboration features (shared history, team spaces, SSO) than commercial agent platforms like Cursor or Devin
⚠️ Recipe/subagent authoring assumes comfort with YAML and command-line workflows
⚠️ Giving an agent broad tool access on your own machine shifts safety responsibility to the operator; guardrails help but do not replace review

Use cases

Multi-file code refactors and PR generationLocal repo Q&A and code reviewResearch assistant with web + filesystem toolsData analysis over local CSVs and SQL databasesScheduled automations via YAML RecipesBrowser-driven scraping and form-filling agentsInternal DevOps triage and log summarizationCustom MCP tool development and testingLocal-first agent for privacy-sensitive workloadsParallel batch processing with subagents

Frequently asked

Is Goose free to use?: Yes, Goose is free and open source under the Apache 2.0 license. Users must provide their own model API keys or run local models via Ollama, with provider costs passed through.
Which AI models does Goose support?: Goose is provider-agnostic and supports 15+ LLM backends, including Anthropic, OpenAI, Google, Azure OpenAI, AWS Bedrock, OpenRouter, and local models via Ollama.
How does Goose integrate with external tools?: It uses the Model Context Protocol (MCP) to attach capabilities from databases, browsers, filesystems, and cloud APIs. There are 70+ ready-made extensions available, and users can write their own.
Can I share or schedule my workflows in Goose?: Yes, the 'Recipes' system allows users to capture a working session as a portable YAML file. This makes workflows reproducible, shareable, and schedulable.
Who is Goose designed for?: It is best for developers and small technical teams who want a fully controllable, local-first agent. It is not intended for non-technical users seeking a plug-and-play chatbot or teams requiring managed SaaS collaboration features.