Open standards and infrastructure for AI agents
Agentic AI Foundation
Agentic AI Foundation (AAIF) stewards open standards and software for developers building AI agents that work across platforms. Its projects cover complementary parts of that infrastructure: Model Context Protocol (MCP) connects agents to tools, data and applications; AGENTS.md gives coding agents repository instructions; and goose provides a runtime for planning and executing work. agentgateway manages routing, policy and observability, while Agent2Agent (A2A) lets independent agents discover one another, delegate tasks and exchange results across frameworks and organizations.
Hosted by the Linux Foundation, AAIF provides neutral governance for these projects. Its leadership includes Manik Surtani, CTO and co-founder, and Mazin Gilbert, Executive Director. A2A uses published agent cards to describe capabilities and connection details, allowing other agents to discover and delegate work.
In August 2026, the foundation reported 247 member organizations, following the addition of 57 over the preceding three months. Alibaba, Visa and Wells Fargo joined as Gold members, extending participation across technology, payments and banking. Separately, AAIF reported that A2A had backing from more than 150 organizations.
2 talks
Newest first
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- Build Systems, Not Code
Start here for a concrete agent example and an explanation of how AGENTS.md, skills, scripts, and sub-agents support maintainable documentation.
Angie JonesAI Engineer World's Fair 2026
- Building an Autonomous Engineering Org
Read this for the connection between repository readiness, improved tools, and delegating bugs and feature implementation through a Slack-based Builder Bot.
Angie JonesAI Engineer World's Fair 2026
Messages from the stage
Make agent workflows recoverable
In Build Systems, Not Code, Jones describes decomposing oversized prompts, using deterministic automation, and maintaining structured, queryable memory. State tracking and crash recovery bring reliability concerns into the agent's design.
Turn tool adoption into delegated delivery
Building an Autonomous Engineering Org describes Block's move beyond widespread but low-impact AI-tool usage. Jones presents a six-stage maturity model and the 1-9-90 rule for cultivating dedicated AI champions.
Affiliations reflect each recorded session, not necessarily current employment.