OpenAI, Anthropic, and Block Launch Agentic AI Foundation to Standardize Autonomous Agents
OpenAI, Anthropic, and Block have launched the Agentic AI Foundation under the Linux Foundation, donating key protocols to establish open standards for AI agents that operate autonomously across tools and data sources. This initiative targets the growing fragmentation in AI agent ecosystems, where proprietary systems risk creating incompatible silos that hinder scalability and adoption. By fostering interoperability, the foundation aims to enable mix-and-match architectures, allowing developers to deploy agents without vendor lock-in.
The Agentic AI Foundation serves as a neutral hub for open source projects focused on AI agents, emphasizing safety patterns, best practices, and shared guardrails for trustworthy operations at scale. Founding members include AWS, Bloomberg, Cloudflare, and Google alongside the primary donors. Governance occurs through technical steering committees, with funding from membership dues ensuring community-driven evolution rather than corporate control.
Anthropic contributes its Model Context Protocol, or MCP, a specification for connecting AI models to external tools and data without requiring vendor-specific adapters. MCP uses a declarative format to define integrations, supporting dynamic tool invocation via standardized APIs that handle authentication and context passing. This protocol addresses current limitations in agent-tool interactions, where custom wrappers increase development overhead by up to 40% in multi-vendor environments.
Block donates Goose, an open source agent framework initially developed for internal tasks like code generation, data analysis, and documentation. Goose employs a modular architecture with pluggable components for task orchestration, leveraging large language models to execute multi-step workflows. Community contributions have already enhanced its extensibility, adding support for over 50 third-party integrations since its inception.
OpenAI provides AGENTS.md, a lightweight markdown file format that embeds instructions for AI coding tools directly in repositories. This file specifies agent behaviors, such as code review scopes or testing protocols, using YAML-like structures parsed by compatible LLMs. AGENTS.md integrates seamlessly with GitHub Actions and similar CI/CD pipelines, reducing setup time for agent-assisted development from hours to minutes.
The foundation’s goals extend to preventing “closed wall” proprietary stacks, drawing parallels to Kubernetes’ success in container orchestration. Success metrics include global adoption rates and implementation counts, with initial focus on U.S.-based enterprises where 65% of AI agent pilots fail due to interoperability issues per Gartner data. Technical working groups will prioritize protocols for agent communication, such as negotiation schemas for multi-agent collaboration.
OpenAI’s Nick Cooper stated, “We need multiple protocols to negotiate, communicate, and work together to deliver value for people, and that sort of openness and communication is why it’s not ever going to be one provider, one host, one company.” Linux Foundation executive director Jim Zemlin added, “By bringing these projects together under the AAIF, we are now able to coordinate interoperability, safety patterns, and best practices specifically for AI agents.”
Anthropic’s David Soria Parra noted on MCP, “The main goal is to have enough adoption in the world that it’s the de facto standard.” Block’s Brad Axen remarked on Goose, “Getting it out into the world gives us a place for other people to come help us make it better.” These contributions position U.S. firms to lead in agentic AI, where domestic data centers process 70% of global inference workloads.
Early adopters, including fintech startups using Block’s Cash App integrations, report 25% faster deployment cycles with standardized agents. The foundation plans quarterly releases of compliance toolkits, aligning with NIST frameworks for AI risk management. As agentic systems power 40% of enterprise automation by 2027, per IDC forecasts, this collaborative effort could accelerate U.S. innovation while mitigating risks from siloed developments.
Broader implications include enhanced security through verifiable agent behaviors, reducing exploit surfaces in distributed systems. Developers gain from open benchmarks for agent performance, such as latency under 200 milliseconds for tool calls. This standardization reinforces open source dominance in AI infrastructure, benefiting U.S. startups reliant on accessible, scalable agent technologies.
