Goose

AI Agents

Block / AAIF

FreeOpen Source

About

Open-source on-machine AI agent from Block (now under the Linux Foundation's AAIF): desktop app + CLI + API, built in Rust, connects to 70+ extensions via MCP

Visit

Share this tool

Twitter LinkedIn Facebook

Our Verdict

Recommended

The free, local-first agent that respects your machine and your wallet

Goose answers a question many developers quietly ask: why pay a platform fee for an agent when the intelligence comes from the model anyway? It runs on your machine, connects to 15+ providers — or reuses the Claude/ChatGPT subscription you already pay for via ACP — and extends through 70+ MCP integrations. The move to Linux Foundation governance removes single-vendor risk. It demands more setup than commercial tools and its ceiling is your chosen model, but as free agents go, nothing else is this complete.

Best for

•Developers wanting a capable agent with zero platform fees
•Privacy-conscious teams keeping code on local machines
•Tinkerers combining local Ollama models with MCP extensions

Consider alternatives if

•You want a top-tier polished coding CLI (→ Claude Code)
•You need a hosted, managed agent platform (→ Factory, Devin)

Try Goose

Supported Platforms

WindowsmacOSLinuxAPI

Available platforms include Windows, macOS, Linux, and API.

Key Features

Open-source AI agent that runs locally: desktop app (macOS/Linux/Windows) plus CLI

Built in Rust for a fast, lightweight on-machine experience

15+ LLM providers, and ACP support reuses your Claude/ChatGPT/Gemini subscriptions

70+ MCP extensions to connect editors, browsers, databases and more

MCP Apps render interactive UI inside conversations

Originally built by Block, now governed by the Linux Foundation's AAIF

Pricing

free

Completely free and open source (Apache 2.0): you pay only your LLM provider, or run fully free with local models via Ollama; ACP support can reuse existing AI subscriptions

Use Cases

Local coding agent that keeps code and data on your machine

Automating engineering tasks from the desktop app or CLI

Connecting an agent to internal tools through MCP extensions

Fully offline agent workflows with Ollama local models

Pros

Truly free: no platform fees, no seat licenses, ever

Local-first design keeps sensitive code off third-party clouds

Vendor-neutral governance under the Linux Foundation AAIF

Desktop app makes agents accessible beyond terminal users

Cons

Output quality depends entirely on the model you connect

No hosted cloud service — you manage setup and updates

Less polished than commercial rivals like Claude Code or Cursor

Governance transition from Block to AAIF is still recent

Latest Update

2026: Goose moves from Block to the Linux Foundation's Agentic AI Foundation (AAIF) for neutral governance, adds ACP support to reuse Claude/ChatGPT/Gemini subscriptions, and ships MCP Apps interactive UI