Agentic Web Stack

Appearance

Agentic Web StackBuild and understand agent-native web systems.

A vendor-neutral reference architecture for AI agents that discover each other, publish capabilities, authenticate, call APIs, use tools, retrieve memory, and collaborate through emerging protocols such as MCP and A2A.

Understand the Stack
Run Local Demo
Explore References
Agentic Web Stack dark mode illustrationAgentic Web Stack bright mode illustration

Stack Map

Discovery, identity, APIs, agent protocols, memory, models, and applications explained as practical technology levels.

Protocol Primers

Clear introductions to Agent Cards, A2A, MCP, OpenAPI, OAuth2, OIDC, vector search, and other agent-web building blocks.

Reference Hub

Curated links to official specifications, origin notes, protocol docs, papers, tools, and practical technologies.

Example Sketches

Each major concept and stack topic includes a small use case plus a representative metadata, API, protocol, or workflow snippet.

Runnable Demo

A local FastAPI service demonstrates Agent Card discovery, scoped JWT access, OpenAPI REST calls, MCP tools, A2A delegation, and a deterministic review workflow.

What This Website Is About ​

Agentic Web Stack is a practical guide to the emerging web of AI agents. It explains how agent-native systems can be made discoverable, secure, callable, interoperable, stateful, and useful across organizational boundaries.

The site is organized for builders who need to answer concrete architecture questions:

  • What metadata should an agent publish so other agents can find it?
  • Which identity layer should protect delegated agent access?
  • Where do REST, GraphQL, OpenAPI, MCP, and A2A fit together?
  • Which memory, model, and application technologies are commonly used?
  • Which standards and protocol links should an architecture start from?

Project Map ​

This site is the knowledge base for the Agentic Web Stack reference architecture.

Three Paths ​

  • Understand the Stack starts with the canonical layer model, architecture variants, and core concepts.
  • Run the Demo executes the Literature Review Assistant and shows discovery, delegation, policy, trace, and review evidence.
  • Explore References collects standards, protocols, timeline history, origins, papers, tools, and glossary definitions.

Canonical Stack ​

Start Here ​

  • Technology Stack explains the seven stack levels and common technologies for each category.
  • End-to-End Demo follows the Literature Review Assistant through every stack layer.
  • Run the Local Demo starts the FastAPI service and executes the deterministic workflow.
  • Layer Model shows the complete reference architecture from discovery to applications.
  • Architecture Flows shows discovery, identity, MCP, A2A, memory, and full-stack diagrams.
  • Architecture Variants shows minimal, enterprise, local/private, and research stack variants.
  • Agent Card explains the machine-readable metadata used for agent discovery.
  • Agent2Agent explains A2A for cross-agent communication and delegation.
  • Model Context Protocol explains MCP for connecting agents to tools and resources.
  • Topic Guides include example use cases and specification sketches for each stack layer.
  • Specifications lists machine-readable examples for the reference scenario.
  • Security and Trust defines trust boundaries and security controls.
  • Decisions records recommended defaults and tradeoffs.
  • Agentic Web Stack Timeline shows how web foundations, model infrastructure, and agent protocols developed over time.
  • Technology Origins tracks when key technologies emerged, who developed them, and which organizations steward them.
  • Glossary defines the core terms used throughout the documentation.

Topic Guides ​

References ​

  • References collect standards, protocols, timeline history, technology origins, papers, and tools.

Page created by Dr. C. Klukas