And why the $200/month price tag for Claude Code might be about to become irrelevant.
If you've been using AI coding tools, you know the drill. You pick one — Claude Code, Cursor, Copilot — and you commit. Monthly subscription. Rate limits that reset every 5 hours. Cloud dependency. No flexibility to switch models mid-session.
Then Jack Dorsey's company Block releases something that makes all of that feel… optional. Goose is a free, open-source AI agent that's changing the economics of AI tooling for developers and businesses alike.
And it's completely free.
Goose is an open-source AI agent built by Block (formerly Square, the fintech company behind Cash App and Square). It was originally built for Block's own engineering teams — they used it internally, measured the impact, and then decided to open-source it under the Apache 2.0 license.
The numbers from internal adoption are hard to ignore:
This isn't a weekend hackathon project. In December 2025, Block donated Goose to the Linux Foundation's Agentic AI Foundation alongside Anthropic's MCP and OpenAI's agents.md. Amazon, Google, Microsoft, and Anthropic are all co-governing it. This is infrastructure.
Goose is a full AI agent — not just a code suggestion tool. It:
There's also a feature called Recipes — YAML workflows with parameters, retry logic, and cron scheduling. Think of it as automation blueprints for your AI agent.
This is where it gets interesting. Here's how Goose stacks up against Claude Code:
| Goose | Claude Code | |
|---|---|---|
| Price | Free (open source) | $20-$200/month |
| Models | Any LLM (30+ providers) | Anthropic only |
| Running | Local + cloud | Cloud only |
| Rate limits | None | Reset every 5 hours |
| License | Apache 2.0 | Proprietary |
| Security testing | Operation Pale Fire red-team | Standard |
The critical difference: Goose is model-agnostic by design. You can start a session with Claude, switch to GPT mid-conversation, and finish with a local Ollama model — all in the same workflow.
In January 2026, Block's own offensive security team ran a three-campaign red team operation against Goose called Operation Pale Fire. They tested it against prompt injection attacks, unicode exploits, and cross-system data leakage scenarios.
They published everything. The attacks, the failures, the fixes. This kind of transparency from a tech company is rare — and it tells you something important about how Block is approaching trust in AI.
Goose uses Anthropic's Model Context Protocol (MCP) — a standardized set of APIs that connect AI agents to the systems where data lives. GitHub, Google Drive, Slack, JetBrains — the list is growing fast.
Block collaborated closely with Anthropic to develop MCP. It's becoming the default standard for AI agent extensibility. If you build with MCP, your agent works with any tool that speaks the protocol.
The release of Goose marks something important: the shift toward open, model-agnostic AI infrastructure.
We're moving from a world where each AI tool locks you into a specific ecosystem, to a world where the framework is open and the model is a choice you make.
The $200/month for Claude Code might still be worth it — the integration quality is slightly better. But knowing there's a free, open, equally capable alternative? That's changing the conversation.
Yes. Goose is completely open-source under the Apache 2.0 license. You only pay for the LLM API calls you make. If you use a free model like Ollama locally, the cost is zero. If you connect your existing Claude Pro or GPT subscription, you pay those existing fees only.
For most developers, Goose offers comparable capabilities with more flexibility. The main difference is that Claude Code has tighter native integration with Anthropic models. If you're fully invested in the Anthropic ecosystem, Claude Code's UX is slightly smoother. But Goose's model-agnostic approach and local execution options make it a serious alternative.
With Goose's multi-model mode, you can use one LLM (like Claude 3.5 Opus) for planning complex tasks and another LLM (like MiniMax or a local Llama) for executing simpler tasks. This lets you optimize for both quality and cost in the same workflow — deep reasoning where needed, fast cheap execution where possible.
Block ran a dedicated red team operation called "Operation Pale Fire" against Goose before release. They published all findings, including attack vectors and fixes. This transparency is unusual in AI tooling and suggests Block takes security seriously. That said, always review any AI agent's capabilities before giving it access to production systems.
Goose is ideal if you want flexibility in LLM choice, prefer local execution for privacy, or are tired of paying high subscription fees for tools that lock you into one provider. It's also excellent if you need to work across multiple LLM providers in a single workflow. If you want the simplest out-of-the-box experience and don't mind the cost, Claude Code is still a strong choice.
Take our free 2-minute scan and get a personalized readiness score.
Take the Free AI Readiness Scan →