OpenAI, Anthropic, and Block have teamed up under the Linux Foundation to launch the Agentic AI Foundation (AAIF), a new body tasked with setting open technical standards for “agentic” artificial intelligence. The group is moving some of today’s most widely used tools for AI agents—Anthropic’s Model Context Protocol (MCP), OpenAI’s AGENTS.md spec, and Block’s Goose framework—into a neutral, open-source home. Together, these projects define how AI agents connect to data, how they are instructed to behave inside codebases, and how developers build production-grade systems on top of them. Major cloud and infrastructure players including AWS, Google, Microsoft, Bloomberg, and Cloudflare are backing the initiative.
Agentic AI refers to autonomous systems that do more than chat: they schedule tasks, execute code, call external tools, and coordinate with other software on a user’s behalf. Up to now, each big vendor has been evolving its own proprietary agent stack, raising fears of fragmentation and lock-in. By pooling MCP, Goose, and AGENTS.md under AAIF governance, the companies are effectively acknowledging that a shared protocol layer is needed if agents are to work reliably across platforms. The Linux Foundation will provide the formal governance, steering committees, and legal framework that helped previous open standards like Kubernetes and Linux become industry defaults.
For developers, MCP is the connective tissue that lets a model call tools, APIs, and data sources in a consistent way, regardless of which AI provider they use. Goose serves as the scaffolding for building local-first, extensible agents that plug into MCP and other tools, while AGENTS.md gives coding agents a single place to read project-specific rules before they start refactoring repositories. Moving these into AAIF means their roadmaps will be shaped in public, with versioned specifications and reference implementations rather than vendor blog posts and one-off SDK updates. That should reduce the risk that a sudden product change by any single company breaks whole fleets of agents.
Implications for smaller markets
For enterprises and governments outside North America, a neutral standard matters as much as the technology itself. Banks, telcos, and logistics firms in markets like Bangladesh often hesitate to bet on proprietary stacks that could be withdrawn or repriced with little notice. A recognized agent standard under the Linux Foundation gives CIOs more confidence that they can mix US clouds, regional data centres, and local open-source projects without rebuilding their automation every year. It also creates room for local vendors to ship MCP-compatible tools—such as regulatory document search, AML checks, or Bangla-language knowledge bases—that can be wired into global agent platforms with minimal glue code.
At the same time, AAIF is also a geopolitical play. A growing ecosystem of Chinese open-source models and toolchains has been offering developers an alternative to US-centric stacks. By turning MCP, Goose, and AGENTS.md into open standards with broad corporate backing, US-aligned firms are trying to ensure that the “grammar” of AI agents is written in a forum they influence. For countries like Bangladesh that buy cloud services from multiple blocs, this could translate into better interoperability but also renewed pressure to align with one standards ecosystem over another.
