As artificial intelligence evolves beyond chatbots toward systems that can take real actions, the Linux Foundation has launched a new initiative aimed at preventing the AI agent ecosystem from fragmenting into incompatible, closed platforms.

The new group, called the Agentic AI Foundation, is designed to serve as a neutral home for open source projects that define how AI agents connect to tools, data and applications. At launch, the foundation is backed by contributions from Anthropic, Block and OpenAI.

Anthropic is donating its Model Context Protocol, a standardized way for models and agents to interact with external tools and data. Block is contributing Goose, its open source AI agent framework. OpenAI is adding AGENTS.md, a simple instruction file that developers can include in repositories to guide how AI coding agents should behave. Together, these components are positioned as core infrastructure for the emerging agent era.

The initiative has attracted broad industry support. Members include AWS, Bloomberg, Cloudflare and Google, signaling a shared interest in common standards that allow AI agents to operate safely and reliably at scale.

According to OpenAI engineer Nick Cooper, protocols function as a shared language that allows different agents and systems to work together without forcing developers to rebuild integrations for every platform. He argues that openness is essential, since the future of AI agents will not be dominated by a single provider or ecosystem.

Linux Foundation executive director Jim Zemlin says the effort is meant to avoid a future dominated by proprietary, closed stacks where agent behavior, orchestration and tool access are tightly controlled by a few vendors. By bringing these projects together, the foundation aims to coordinate interoperability, safety patterns and best practices specifically for AI agents.

Block’s involvement may seem unusual for a fintech company best known for Square and Cash App, but the company sees Goose as proof that open source agents can operate at scale. According to Block AI tech lead Brad Axen, thousands of engineers already use Goose weekly for tasks like coding, data analysis and documentation.

Open sourcing Goose also allows external contributors to improve the framework, with those enhancements flowing back into Block’s own engineering efforts. Donating the project to the Linux Foundation further positions Goose as a practical example of how shared standards like MCP and AGENTS.md can work together.

Anthropic is making a similar move at the protocol level by placing MCP under Linux Foundation governance. The goal is to drive broad adoption so developers can build integrations once and reuse them across different tools and clients. MCP co creator David Soria Parra says this kind of neutral infrastructure benefits the entire ecosystem by eliminating the need for countless custom adapters.

Governance is a central reason the Linux Foundation created a separate umbrella for agent technologies. While the organization already hosts major projects such as PyTorch, Ray and Kubernetes, the Agentic AI Foundation is focused specifically on agent orchestration, interoperability and shared safety patterns.

The foundation is funded through a directed fund model, with companies contributing via membership dues. Zemlin emphasizes that funding does not translate into control, as technical steering committees set project direction and no single member can dictate outcomes.

Whether the Agentic AI Foundation becomes true infrastructure or simply another industry alliance remains an open question. Zemlin says early signs of success will include real world adoption and consistent use of shared standards by vendor built agents across the industry.

Cooper adds that long term success depends on continuous evolution. He says the protocols must keep improving and incorporating new input rather than remaining static.

There is also the possibility that one implementation could become dominant due to faster adoption or broader usage. Zemlin notes that open source history shows this is not necessarily a problem, pointing to Kubernetes as an example of a project that rose to prominence through merit rather than vendor control.

For developers and enterprises, the near term benefits are straightforward. Shared standards reduce the need for custom connectors, create more predictable agent behavior and simplify deployment in security sensitive environments.

The broader ambition is more transformative. If tools like MCP, AGENTS.md and Goose become foundational infrastructure, the AI agent landscape could move away from closed platforms toward an open, interoperable ecosystem similar to the one that enabled the modern web.

Source