Coding agent (CLI) · Agentic AI Foundation
goose
General-purpose open-source agent with a desktop app and a CLI, not limited to code. Extends through MCP servers, supports recipes and subagents, and works with many cloud and local models.
goose review
goose is an open-source AI agent that ships as both a desktop app and a CLI for Windows, macOS and Linux. It is free under Apache-2.0. The repository now sits under aaif-goose and the project is part of the Agentic AI Foundation. We read the vendor's pages and repository in October 2026. We did not run an independent test.
Who it is for
Users who want an open-source CLI coding agent with a graphical option, MCP extensions and a free choice of model, including local ones. goose is a general agent, not only a code tool.
What stands out
The desktop app and the CLI share the same agent. Install the CLI with a curl script from the GitHub releases, or get the desktop app by download or `brew install --cask block-goose`. For non-interactive use, `goose run -t` and instruction files cover scripts. Release v1.56.0 appeared on 23 October 2026.
Extensions follow the MCP standard, and the docs list more than 70. A documented subagent feature lets goose hand parts of a job to other agents in parallel. The docs also describe permission modes and prompt-injection detection.
Goose works with more than 15 providers, including Anthropic, OpenAI, Google, Azure, Bedrock and OpenRouter. Local runners such as Ollama, LM Studio and Docker Model Runner also work. Through ACP providers, goose can reuse an existing Claude, ChatGPT or Gemini subscription.
Price and licence
goose is free, checked on 28 October 2026. The licence is Apache License 2.0, which is open source. Model cost comes from whatever you plug in. That can be your own API keys, a subscription you already pay for through ACP, or a local model.
Where it falls short
- It is positioned as a general agent. Coding-specific features such as a repo map or LSP are less central than in code-only tools.
- We found no built-in command sandbox in the docs. Isolation comes from a third-party MCP, container-use, shown in a 2025 blog post.
- The repo moved from block/goose to aaif-goose/goose, so older links and tutorials may use the previous name.
- Roaming agents, which reach a machine remotely over p2p, are marked experimental and are not in released binaries.
- We did not verify git worktree support, a mobile app or team sharing.
Alternatives
For a wider comparison, see Claude Code alternatives. Other open-source agents with MCP are OpenCode and Qwen Code.
Facts: platforms, price and licence
- platforms
- macOSWindowsLinuxShips both a desktop app (macOS, Linux, Windows) and a CLI, plus an API. Built in Rust. Repo moved from block/goose to aaif-goose/goose; the project is described as part of the Agentic AI Foundation (AAIF) at the Linux Foundation.
- price
- Freegoose is free (Apache-2.0, seen 2026-10-28). Model cost comes from the provider you plug in: your own API keys, existing Claude/ChatGPT/Gemini subscriptions through ACP providers, or local models (Ollama, LM Studio, Docker Model Runner).
- source
- Open sourceLicence: Apache-2.0
- github
- aaif-goose/goose 55.8k stars, seen 2026-10-28
goose features
- ✓ yes MCP
- ✓ yes Multi-agent
- ✓ yes Headless / CI
- ? not verified Sandbox
- ✓ yes Local models
- ✓ yes Your own API key
- ? not verified Git worktrees
- ? not verified SSH
✓ documented · – absent or not in its feature list · ? not verified
goose strengths
- Desktop app and CLI share the same agent; `goose run -t` and instruction files cover non-interactive use.
- Extensions use the MCP standard (70+ listed) and a documented subagent feature allows delegated parallel work.
- Permission modes and prompt-injection detection are documented.
- Wide provider list, including local runners and subscription reuse through ACP.
goose weak points
- Positioned as a general agent, so coding-specific features (repo map, LSP) are less central than in code-only tools.
- No built-in command sandbox found in the docs; isolation comes from a third-party MCP (container-use) described in a blog post.
- Governance and repo location changed (block/goose to aaif-goose/goose), so older links and tutorials may point to the previous name.
- Roaming agents (remote reach over p2p) is experimental and not in released binaries.
Who goose is for
- best for
- Users who want an open-source agent with both a GUI and a CLI, MCP extensions and the freedom to choose any model, including local ones.
- look elsewhere if
- Anyone who needs a hardened, sandboxed execution environment built in.
Compare goose
Lists that include goose
Sources and what we could not verify
Checked 2026-10-28. Prices and features change; the vendor's page is the authority.
- goose-docs.ai — official docs
- github.com/aaif-goose/goose — source Apache-2.0, stars, README: desktop + CLI, platforms, AAIF, 15+ providers, MCP
- github.com/aaif-goose/goose/blob/main/LICENSE — giấy phép Apache License 2.0
- github.com/aaif-goose/goose/blob/main/documentation/docs/getting-started/installation.md — install CLI and desktop, macOS/Linux/Windows
- github.com/aaif-goose/goose/blob/main/documentation/docs/getting-started/providers.md — Ollama, LM Studio, Docker Model Runner, API keys, ACP subscriptions
- github.com/aaif-goose/goose/blob/main/documentation/docs/guides/running-tasks.md — goose run -t headless
- github.com/aaif-goose/goose/blob/main/documentation/docs/tutorials/subagents.md — subagents
- github.com/aaif-goose/goose/blob/main/documentation/docs/guides/roaming-agents.md — roaming agents experimental, not in released binaries
Not verified
- goose-docs.ai homepage returned little text to my fetch; docs read from the GitHub repo's documentation/ folder instead
- sandbox: no built-in sandbox found; container-use is a third-party MCP shown in a 2025 blog post (null)
- git_worktrees, mobile_app, team_sharing not verified (null)
- vendor: Block originally created goose; README now says it is part of AAIF at the Linux Foundation