As artificial intelligence rapidly shifts from passive chatbots to autonomous systems capable of making decisions, completing tasks, and operating across digital environments, the race to build AI “agents” is escalating fast. However, with each company building its own tools, protocols, and compatibility layers, the agent landscape risks splintering into exactly what the internet once sought to avoid: a maze of fragmented standards and closed systems.
To prevent this from happening, the Linux Foundation has launched the Agentic AI Foundation (AAIF), a new open-source umbrella project designed to standardise the development, connection, and deployment of AI agents. The initiative arrives with heavyweight backing: OpenAI, Anthropic, and Block are donating key technologies to form the foundation’s core infrastructure.
With additional support from AWS, Google, Cloudflare, and Bloomberg, the AAIF is shaping up to be the first industry-scale alignment effort for the next era of AI.
Setting Common Ground for the Agent Era
At its core, the AAIF aims to create the “basic plumbing” required for a world where AI agents can safely, reliably, and efficiently work across different platforms. The goal is simple yet ambitious: prevent the future of AI agents from becoming controlled by a handful of proprietary products.
To achieve that, the foundation is establishing shared protocols, the equivalent of a universal language, that allows different models, tools, and agents to interact without custom integration work at every turn.
The founding members have contributed three foundational technologies that represent the first layer of this vision:
1. Anthropic’s MCP (Model Context Protocol)
MCP offers a standardised way for agents and models to access data, tools, and external systems. Instead of every developer building bespoke connectors, MCP formalises the pathway between an AI system and the resources it relies on.
By donating MCP, Anthropic ensures the protocol can’t be captured by a single vendor, a meaningful step toward building neutral infrastructure for AI development at scale.
2. OpenAI’s AGENTS.md
AGENTS.md is a lightweight but pivotal contribution. It’s a simple instruction file that tells AI coding tools how to behave within a given repository. If broadly adopted, it could function as a universal configuration layer for agent-assisted development across the open-source ecosystem.
3. Block’s Goose Framework
Block’s “Goose” is an open-source agent framework already used internally by thousands of its engineers for tasks ranging from documentation to data analysis.
By handing Goose to the AAIF, Block positions itself as a working proof of concept for open agent tooling and benefits from community-driven improvements that flow back into its systems.
Why Block’s Move Matters
Block isn’t typically mentioned in the same breath as OpenAI or Anthropic when discussing AI infrastructure, but its contribution is strategically important.
According to Block AI lead Brad Axen, open tooling is essential if the ecosystem is going to remain competitive and prevent agent technology from consolidating under closed, proprietary platforms.
Goose already integrates smoothly with MCP and AGENTS.md, making Block a live case study of the agent-interoperability vision the AAIF wants to scale.
For Block, the incentives are clear: more contributors, more battle-testing, and more alignment with the industry’s emerging standards.
A Push for True Interoperability
The larger ambition behind the AAIF mirrors how the early internet evolved.
Jim Zemlin, Executive Director of the Linux Foundation, emphasised that without shared standards, companies risk creating siloed agent ecosystems, each with its own orchestration rules, tool integrations, and safety protocols.
The AAIF aims to establish a common foundation for:
- interoperability
- safety patterns
- best practices
- scalable, predictable agent behaviour
For developers and enterprises, the benefits are immediate:
- fewer custom connectors to build
- consistent agent behaviour across environments
- easier adoption of secure deployment patterns
- improved compatibility across vendors and frameworks
Ultimately, a standardised agent stack reduces friction, accelerates development cycles, and lowers the barrier to enterprise-grade adoption of autonomous AI.
Governance Without Gatekeeping
The AAIF will be funded through a “directed fund” supported by member contributions. However, technical decisions will be shaped by independent steering committees, meaning no single company, not even the largest contributors, can dictate the foundation’s roadmap.
This governance model mirrors the Linux Foundation’s existing approach: neutral oversight with community-balanced direction.
Success now hinges on adoption. If AI vendors, developers, and enterprises embrace these standards, the AAIF could become to AI agents what the Internet Engineering Task Force (IETF) was to the modern web: a blueprint that defines how everything works together.
The Road Ahead
The Agentic AI Foundation marks one of the first serious industry-wide attempts to bring order to an emerging AI wave before chaos sets in. With contributions from three of the most influential players in AI, and backing from the world’s largest cloud and infrastructure providers, the initiative has momentum, credibility, and a clear mission.
If it succeeds, AI agents may evolve not as siloed, proprietary systems, but as interoperable, secure, and universally compatible computing partners.
And for the developers building the next generation of AI, that’s a future worth rooting for.
