AAIF
The Agentic AI Foundation
A Linux Foundation consortium co-founded by Anthropic, OpenAI, and Block. AAIF governs the open standards that power the Internet of Agents -- starting with MCP, the protocol that connects AI agents to tools.
What is AAIF?
The Agentic AI Foundation (AAIF) is an open-governance body under the Linux Foundation created to steward the standards that agentic AI systems depend on. Founded in early 2026, it brings competing AI labs -- Anthropic, OpenAI, Block (formerly Square), and others -- under a neutral governance structure to ensure that core protocols remain open, interoperable, and vendor-neutral.
AAIF's founding principle: no single company should control the protocols that agents use to talk to tools, to each other, or to humans. Just as the Linux Foundation governs Kubernetes, CNCF, and Node.js, AAIF governs the agent infrastructure layer.
Founding members
Governance structure
Linux Foundation | Open governance | Vendor-neutral
Agent-to-tool communication
capability description
agent framework
instructions standard
Why AAIF matters
Before AAIF, each AI lab developed its own protocols independently. Anthropic created MCP. Google created A2A. Cisco/AGNTCY created SLIM. This fragmentation risked creating the same vendor lock-in that AAIF's members publicly oppose.
- Neutral governance -- No single company can veto protocol changes or steer standards toward proprietary advantage. Decisions require broad consensus.
- IP protection -- Contributor License Agreements (CLAs) ensure that contributions remain open. No patent traps, no license changes.
- Interoperability testing -- AAIF maintains conformance test suites. An MCP implementation that passes the tests is guaranteed to work with any other compliant implementation.
- Ecosystem coordination -- When MCP, A2A, and SLIM need to interoperate, AAIF provides the forum for cross-protocol alignment.
The projects
- MCP -- The foundational protocol. Lets AI agents connect to tools, data sources, and services. Originally created by Anthropic, now community-governed under AAIF.
- AI Card -- A standardized format for describing AI agent capabilities, similar to Agent Cards but with richer metadata for trust and compliance.
- goose -- An open-source AI agent framework (originally by Block) that provides a reference implementation for AAIF-governed protocols.
- AGENTS.md -- A standard for repository-level agent instructions, similar to ROBOTS.txt but for AI agents. Tells agents what they can and cannot do in a codebase.
How MeetLoyd participates in AAIF
MeetLoyd is an active participant in the AAIF ecosystem with full production implementations:
- MCP 100% implemented -- MeetLoyd's agent runtime is built on MCP from the ground up. Every tool connection, every data source integration, every agent-to-tool interaction uses MCP natively.
- Contributing to standards -- MeetLoyd contributes real-world implementation feedback from enterprise deployments across financial services, healthcare, and legal verticals.
- Multi-protocol support -- Beyond MCP, MeetLoyd implements A2A for agent collaboration, SLIM for cross-org federation, and SPIFFE for agent identity.
- Governance layer -- Where AAIF provides the protocols, MeetLoyd adds the enterprise governance that regulated industries require: audit trails, compliance packs, spending controls, and identity verification.