From Internal Tool to Open Standard: How Goose Found a Home at the Linux Foundation

The Governance Gap in Open Source AI Agents

When Block—the fintech powerhouse behind Square and Cash App—open-sourced its internal AI coding agent Goose, the move was celebrated as a generous gift to the developer community. Under a permissive license, any organization could adopt and modify Goose for their own workflows. But soon, a critical flaw emerged: Block retained full ownership of Goose’s trademarks. This subtle yet significant control created headwinds for enterprise adoption, as companies hesitated to build critical infrastructure on a project whose brand and direction remained under a single corporate entity.

From Internal Tool to Open Standard: How Goose Found a Home at the Linux Foundation — Source: thenewstack.io

“Goose wasn’t entirely free and open, as Block still owned its trademarks, hampering enterprise adoption,” explained Manik Surtani, former head of open-source at Block and now CTO and co-founder of the Agentic AI Foundation (AAIF). In a conversation at an MCP developer event, Surtani revealed that despite Goose’s rapid early adoption after its public release, the lack of transparent governance became a major barrier. “The biggest issue was a lack of transparency in governance for the Goose project,” he said. “Something had to change if Goose would be able to fully unfurl its wings and soar.”

Why Traditional Open Source Falls Short for AI Tools

Open-source projects typically live or die by their governance model. When a single company owns the trademarks, it can unilaterally change licensing terms, redirect development, or even discontinue support. For enterprises building agentic AI systems—where reliability and continuity are paramount—such risks are unacceptable. Goose needed a neutral home where the community, not just Block, would steward its future.

The “Headwinds” Block Encountered

After the initial buzz of the open-source release, the Goose team at Block noticed a plateau in contributions and slower enterprise uptake. Surtani detailed two main obstacles:

Trademark dependency: Companies feared investing in a project whose brand and direction could change overnight.
Lack of community trust: Developers wanted assurance that the project’s roadmap would be shaped by consensus, not corporate priorities.

“We saw that the project had huge potential, but enterprise buyers were asking, ‘Who really owns this? Where is the steering committee?’” Surtani noted. The answer was clear: Goose needed a foundation.

The Solution: Creating a Home Under the Linux Foundation

Block began exploring foundation options. After discussions with the Model Context Protocol (MCP) team at Anthropic and other key players in the AI agent ecosystem, a natural partnership emerged. Rather than joining an existing foundation, the group decided to launch a new entity dedicated to agentic AI: the Agentic AI Foundation (AAIF). This foundation would live under the umbrella of the respected Linux Foundation, known for its neutral governance of projects like Kubernetes and the Cloud Native Computing Foundation.

The move addressed the governance deficit directly. Under AAIF, Goose’s trademarks would be transferred to the foundation, ensuring no single company could control the project. A governance board with representation from multiple organizations—including Block, Anthropic, and early adopters—would oversee development. Surtani emphasized that this structure was designed to build trust: “Enterprises now see that the project is vendor-neutral and backed by a legal framework that protects their investments.”

The Trio of Tools That Launched AAIF

The Agentic AI Foundation didn’t start with Goose alone. To create immediate momentum, the founding group included three core projects:

Goose – Block’s AI coding agent, now fully community-governed.
Model Context Protocol (MCP) – An open standard for connecting AI agents to external data and services, originating from Anthropic.
Agents.MD – A framework for building and composing AI agents.

“Launching the foundation with the trio of tools was partially due to expediency,” Surtani admitted. “We wanted to get the group up and running quickly, and these three projects already had active communities and complementary visions.” The foundation’s charter explicitly welcomes additional projects; any agentic AI tool that meets governance criteria can apply for membership.

What This Means for the Future of AI Agents

The transfer of Goose and its sibling projects to the Linux Foundation signals a maturing of the AI agent ecosystem. Just as Kubernetes flourished under neutral governance, the AAIF model could accelerate enterprise adoption of AI coding agents by removing the fear of vendor lock-in. For developers, it means a stable, community-driven platform whose trademarks and roadmap belong to everyone.

Surtani expects the foundation to grow: “We’re already in talks with several large tech companies about contributing additional tools. The goal is to make AAIF the neutral hub for all things agentic AI.” For now, Goose has found a permanent home—and a governance model that lets it truly fly.

For more on how Block successfully deployed AI agents across its entire workforce, see our related coverage: How Block Got 12,000 Employees Using AI Agents in Two Months.

Tags: