Explore the Tracks

The Agentic Commerce track covers what happens when AI agents start shopping, selling, and paying on behalf of people and businesses. Sessions explore the protocols and infrastructure emerging to let agents transact (e.g. discovery, cart, checkout, and payment) and the practical realities of running commerce through an agent instead of a human.

Who Should Attend?

Engineers and product leaders building checkout, payments, or commerce integrations for AI agents; platform teams evaluating agentic payment protocols; merchants and platforms trying to understand what “agent-ready commerce” actually requires.

What will you learn?

You’ll come away understanding the protocols and patterns shaping agent-driven commerce, what it takes to let an agent safely complete a purchase on someone’s behalf, and the risks and open questions that come with money moving through agents instead of people.

The Agentic Engineering track covers the craft of building agents. Sessions cover the harnesses, workflows, and engineering practices that turn a model into a working agent system. Expect deep dives into coding harnesses and agentic software factories, agent skills and how to ship them safely, context and memory design, cost and compute efficiency, and the everyday engineering decisions behind agent systems, including when an agent is the right tool and when it is not.

Who Should Attend?

Engineers who build agents and the harnesses that run them; teams who design agentic workflows; anyone who makes day to day engineering decisions about how an agent should be built and deployed.

What Will You Learn?

You’ll learn practical patterns for building agent harnesses, workflows, and skills. You will learn how teams manage agent context, memory, and cost as their systems grow. You will hear how engineers decide when an agent is the right choice for a task and when a workflow or a service does the job better, based on real projects.

The Building Reliable Agent Systems track explores what it takes to keep agents trustworthy once they’re doing real work in production. Sessions cover architecture patterns that make agents less brittle, guardrails and sandboxing for safe autonomy, observability into what agents are actually doing, and the recurring failure modes teams run into once agents move past the demo stage.

Who Should Attend?

Engineers operating agents in production, teams responsible for agent uptime and trust, anyone who has watched an agent fail in a way that was hard to explain afterward.

What Will You Learn?

You’ll learn practical patterns for making agents more predictable and recoverable in production, how teams design guardrails, sandboxes, and observability into agent systems, and the common ways agents fail in the real world along with what actually fixes them.

The Enterprise Adoption in Practice track is grounded in real deployments at real companies, covering production case studies, cost and access control, regulated-industry constraints, and the organizational realities that shape whether an agent initiative succeeds or stalls.

Who Should Attend?

Engineering leaders and decision-makers driving enterprise AI adoption; teams operating agents under compliance, security, or cost constraints; anyone translating agentic AI pilots into durable production systems.

What Will You Learn?

You’ll learn what actually happens when agents hit production traffic and real users inside large organizations, how teams manage cost, identity, and access at enterprise scale, and the practical and organizational lessons behind successful enterprise AI adoption.

The Evals & Testing track tackles the hard problem of knowing whether an agent actually works. Sessions move past simple pass/fail checks toward richer ways of measuring agent quality, catching regressions before they ship, and building testing practices suited to systems that reason and act rather than just respond.

Who Should Attend?

Teams building eval pipelines, QA and platform engineers responsible for agent quality, anyone shipping prompt or agent changes without a safety net today.

What Will You Learn?

You’ll learn why traditional testing approaches fall short for agentic systems, how teams build regression suites, evaluation pipelines, and release gates for agents, and practical ways to measure agent quality and safety beyond a simple pass or fail.

The Human-Agent Collaboration track examines the interface between people and agents including how agents should communicate, ask for approval, and share context with the humans working alongside them. It also explores what changes in design, process, and trust once an agent becomes a genuine collaborator rather than just a tool.

Who Should Attend?

Product and UX engineers designing agent-facing interfaces; teams building approval, oversight, or trust workflows into agent products; anyone rethinking what collaboration looks like when one of the collaborators is an agent.

What Will You Learn?

You’ll learn patterns for keeping humans meaningfully in the loop on agent decisions and actions, how interfaces and workflows are evolving to support agents as collaborators, and what separates genuinely productive human-agent collaboration from tool use alone.

The Interoperability & Standards track covers the protocols and open standards that let independently built agents, tools, and platforms work together safely. Sessions span agent identity and delegation, access and security boundaries, and the evolving standards that hold the ecosystem together.

Who Should Attend?

Protocol designers and standards contributors, security engineers hardening agent access boundaries, teams building agent discovery, identity, or delegation systems across organizational or vendor boundaries.

What Will You Learn?

You’ll learn how agent identity, delegation, and access control are being standardized across the ecosystem, why security boundaries for agents have to assume things will go wrong rather than just hope they won’t, and where key protocols are heading next.

MCPCon, the MCP-focused track within AGNTCon, centers on the protocol’s latest release — the most significant shift in MCP’s history. Sessions cover the move to a stateless protocol core, the new extensions framework (including Tasks and MCP Apps), authorization hardening, and what it takes to migrate real servers, clients, and SDKs onto the new spec. Talks bring hands-on migration experience and production lessons.

Who Should Attend?

MCP builders, maintainers, and adopters; anyone with a server, client, or SDK that needs to conform to the new spec; security teams assessing what the authorization changes mean for their deployments.

What Will You Learn?

You’ll learn about MCP’s shift to a stateless protocol and how that reshapes session handling, request identity, and gateway routing. You’ll see what the new extensions framework and Tasks make possible, understand the authorization changes replacing older client-registration approaches, and hear real migration experience from teams moving servers, clients, and SDKs onto the new spec rather than just a summary of what shipped.

The Multi-Agent + Distributed Systems track treats fleets of cooperating agents as the distributed systems problem they actually are. Sessions cover coordination, orchestration, and failure patterns that emerge once you have more than one agent working together (or against each other) at scale.

Who Should Attend?

Engineers designing systems of multiple cooperating or competing agents, distributed systems practitioners moving into agentic workloads, SRE and platform teams running agent fleets in production.

What Will You Learn?

You’ll learn how teams design coordination and orchestration for systems of multiple agents, the distinctive failure modes that show up once agents have to cooperate or compete, and lessons from running multi-agent systems in real, high-scale production environments.

The Open Source Community + Ecosystem Health track looks at what open source looks like now that agents are active participants writing code, opening PRs, and working inside repos alongside human maintainers. Sessions cover how projects adapt their conventions and documentation for agent contributors, what changes when a human and an agent are building side by side, and the sustainability questions that come with an ecosystem where agents increasingly do the work.

Who Should Attend?

Open source maintainers and contributors adapting their projects for agent contributors, engineering leaders thinking about what agent-driven contribution means for project health, anyone rethinking maintainer conventions now that agents are working in their repos.

What Will You Learn?

You’ll learn how projects are adapting conventions, documentation, and contribution norms for agent contributors, what it actually looks like when a human and an agent collaborate on the same codebase, and the sustainability questions facing an ecosystem where agents are doing a growing share of the contribution work.

The Open Source Tools track covers open source options for the practical infrastructure layer builders need underneath their agents such as routing, sandboxing, self-hosting, and the tooling that makes agent systems observable, portable, and cost-effective to run.

Who Should Attend?

Platform and infrastructure engineers building the plumbing under agentic systems; teams choosing between managed and self-hosted model infrastructure; anyone standardizing on skills, sandboxes, or agent gateways.

What Will You Learn?

You’ll learn how teams are architecting the infrastructure layer beneath their agents, the tradeoffs between managed and self-hosted approaches, and open tooling patterns that make agent systems easier to run, secure, and scale.