I’m writing this from IETF 126 in Vienna, where there is a lot of Agentic AI work percolating this week. Before I get into that — Agentic Internet Workshop #3 is November 6 at the Computer History Museum in Mountain View, with an Interop Day on November 5. Register on Eventbrite or learn more at agenticinternetworkshop.org.
Now, here’s what’s happening at IETF that makes AIW #3 more important than ever. Three of the five Birds of a Feather sessions at this IETF are focused on AI agents:
- agentproto — a proposal to charter the first IETF working group for agent-to-agent and agent-to-tool communication protocols
- DAWN — Discovery of Agents, Workloads, and Named Entities, tackling how agents find each other at scale
- DMSC — Dynamic Multi-Agent Secured Collaboration, exploring how agents coordinate across trust boundaries
On top of that, the WIMSE working group on workload identity is meeting, AIPREF (AI Preferences) is holding a session, OAuth has two days of meetings, and there are also many side meetings about Agentic AI including one for MCP.
The message is clear: the identity, authorization, and trust layer for AI agents is now a first-class internet infrastructure problem. The IETF — the organization that standardizes the protocols that run the internet — is treating it that way.
AIW as A Neutral Field for AI Protocol Ecosystem
The agentic AI landscape is exploding in every direction at once. The AAIF (Agentic AI Foundation at the Linux Foundation) now stewards several major agent protocols including MCP (originally from Anthropic), A2A (originally from Google), and Goose. Other protocols are emerging from open communities like ANP and AITP. Standards work is happening at IETF, W3C, OpenID Foundation, DIF, OWASP, ITU SG17, the Cloud Security Alliance, and the Advanced AI Society. Agent frameworks are proliferating — LangChain, CrewAI, n8n, and many more. Companies large and small are building agent infrastructure. Researchers at Stanford, Oxford, and elsewhere are studying the implications.
All of these communities are working on pieces of the same puzzle — but they don’t always talk to each other. When they do, it’s often in formal settings that don’t leave room for the messy, honest, exploratory conversations that lead to real breakthroughs.
AIW exists to be the coherent, neutral field where all of these communities come together. We’re actively reaching out to every context where agentic AI work is happening — protocol teams, standards bodies, open source frameworks, enterprise platforms, academic researchers, and policy thinkers — because the problems we’re trying to solve can’t be solved by any one community alone. Agent identity, authorization, trust, discovery, and governance require cross-pollination that only happens when you put people from different worlds in the same room and let them self-organize around the problems that matter most.
That’s what the unconference format does. No pre-set agenda, no keynotes, no vendor pitches. Just the people doing the work, in a professionally facilitated space, building on each other’s knowledge in real time.
AIW #3 — November 6
Agentic Internet Workshop #3 is November 6, 2026 at the Computer History Museum in Mountain View, California. It follows IIW #43 (November 3-5) at the same venue. Registration is live.
We’re also running an Interop Day on November 5, in parallel with Day 3 of IIW. We experimented with this at AIW #2 and we’re bringing it back.
Andor Kesselman, Phil Windley, and I founded AIW in 2025 to bring the same unconference format that has made IIW successful for over 20 years into the agentic protocol space. IIW is where OpenID Connect and OAuth were incubated — protocols now used between 10 and 15 billion times a day. The agentic internet needs the same kind of cross-community design work, and that’s what AIW provides.
Growing the Community — Our New Conveners
I’m excited to annouce our community conveners who are helping shape AIW #3 and activate the community between events:
- Mike Prince
- Ken Adler
- Sarah Cecchetti
- Clawdrey Hepburn) — our AI Agent ambassador
You can see their bio’s on the who’s coming page.
What Happened at AIW #2
AIW #2 was held May 1, 2026 at the Computer History Museum. Over 120 people participated in 44 sessions across four breakout rounds and a lunch session. Protocol creators from MCP, A2A, ANP, and AITP were in the room alongside people active in IETF, W3C, OpenID Foundation, and DIF.
Sessions covered:
- Agent discovery — “How Does My Agent Find Your Agent? A2A MCP DID”
- Delegated authorization and OAuth for agents — AAuth, OAuth Actor Profile
- Agent-to-agent protocol design — ANP cross-domain communication
- Trust boundaries and fiduciary AI — Fiduciary AI and Decentralized Trust Graphs
- Governance — GovOps: How Enterprises Can Govern AI Agents
- Human-in-the-loop controls — Buck Stops with a Human, A NameSpace that Puts the Human-In-the-Loop
- Agent identity — Convince Me I’m Wrong: AI Agent Identity is Useless, Cross-Organizational AI Agent Identity
- Payment and credentials — KYAPay: Agentic Identity and Payment Credentials
The full Book of Proceedings is available and documents all 44 sessions.
Topics Shaping Up for AIW #3
We already have suggested topics for AIW #3, including:
- Security and authority frameworks for agent chains — revocable and auditable authority across MCP and A2A connections
- Agent identity, authority, memory, and intent layers
- Behavioral coherence as a trust signal
- Human authority and execution-time consent/revocation
- Organizational identity for agents
- A2A and OAuth support for agentic workflows
- How MCP and A2A are converging on identity and mandate
- Trust signals that travel with agents without enabling surveillance
Of course, AIW uses Open Space Technology — the agenda is created live the morning of the event. These are starting points, not a fixed program. If something lands in the agentic AI world in October, it can be on the agenda November 6.
Why This Matters Now
What I’m seeing at IETF 126 confirms what we’ve been saying since we launched AIW: the identity layer for agents can’t be an afterthought. The protocols being proposed here this week — for agent communication, discovery, and collaboration — all need identity, authorization, and trust baked in from the start. And that design work needs to happen across communities, not within any single standards body.
Join Us
Register for AIW #3 — November 6, 2026, Computer History Museum, Mountain View.
Register for IIW #43 — November 3-5, same venue. Many people come for the full week.
If you’re working on agent protocols, identity, authorization, or governance — or if you know someone who is — please share this post and point them to agenticinternetworkshop.org.
Sponsors for AIW #2 included AWS, Skyfire, and JLINC Labs. If your organization is interested in sponsoring AIW #3, get in touch: kaliya@identitywoman.net
Leave a Reply