Agent ExplorerBeta

Introduction

Agent Explorer is a web-based testing and validation toolkit for the Agent-to-Agent (A2A) protocol and MCP (Model Context Protocol). It helps developers test real agents and MCP servers, build configurable mocks, and debug protocol interactions — all from one interface.

Agent Explorer — test A2A agents and MCP servers you can trust

What is A2A?

The Agent-to-Agent (A2A) protocol is an open standard that enables AI agents to communicate with each other. It defines how agents discover each other's capabilities, exchange messages, and coordinate on tasks.

What Agent Explorer Does

Agent Explorer provides a full suite of tools across three main areas:

Testing A2A Agents

  • AgentCard Discovery & Validation — Fetch and validate agent cards from any A2A-compliant agent
  • Payload Builder — Build and test A2A message payloads interactively
  • Conversation Testing — Send messages and track multi-turn conversations
  • Task State Tracking — Monitor task states and transitions
  • Push Notification Webhooks — Receive real-time task updates via webhooks with full delivery observability
  • Request/Response Logging — Inspect raw HTTP requests and responses
  • OAuth 2.0 Support — Authenticate with secured agents

Testing MCP Servers

  • Real MCP Playground — Connect to any Streamable HTTP MCP server and browse its tools, resources, and prompts
  • Capability Invocation — Invoke tools, read resources, and render prompts directly with a structured form — no JSON-RPC required
  • MCP OAuth 2.1 — Full PKCE S256 + Protected Resource Metadata discovery for OAuth-secured MCP servers

Building Mock Agents & MCP Servers

  • Custom Mock Agents — Create fully configurable A2A agents via a 3-path creation flow (URL, file, or blank) and edit them in the Configuration tab
  • Mock MCP Studio — Define tools, resources, and prompts with per-capability visual flows
  • Visual Flow Builder — Design response logic using a node-based visual editor (works for both A2A and MCP)
  • Simulation Features — Simulate errors, delays, and custom task states
  • Skill Routing — Route messages to different response flows by skill (manual, pattern, or LLM-based)
  • LLM Integration — Connect OpenAI-compatible models for intelligent mock responses

Collaboration

  • Teams — Organise agents and MCP servers by team
  • Workspaces — Group A2A agents and MCP servers together for project organisation
  • Shared Access — Team members (owner, admin, member) share resources

Who Is It For?

Agent Explorer is designed for:

  • Agent Developers building A2A-compliant agents
  • MCP Server Developers building tools, resources, and prompt servers
  • Integration Engineers connecting systems via A2A or MCP
  • QA Teams testing agent and MCP server implementations
  • Anyone learning the A2A or MCP protocol