Structured and standardized
Fixed, machine-parseable fields for identity, interfaces, capabilities, skills, and security, defined by the open A2A specification.
Agent Card for Agent2Agent (A2A)
Create, validate, and publish A2A Agent Card JSON for AI agent discovery.
{
"name": "Customer Support Agent",
"description": "A2A agent for support tasks",
"version": "1.0.0",
"supportedInterfaces": [{
"url": "https://api.example.com/a2a",
"protocolBinding": "JSONRPC",
"protocolVersion": "1.0"
}],
"skills": [{ "id": "resolve-support-request", … }]
}What is an Agent Card?
An Agent Card is a standardized JSON document that declares an AI agent's identity, capabilities, skills, and access requirements. It is the README and business card of the agent era — published at a well-known URL so people and machines can understand what an agent does before calling it.
Readable by humans and machines
One JSON document serves integrators reading documentation and orchestrators routing tasks automatically.
Versioned and dynamic
Discovery metadata evolves with the agent: the version field and JWS signatures let clients detect and trust changes.
Explicit boundaries
Authentication requirements and skill limits are declared up front, before any call is made.
Why agents need Agent Cards
Thousands of agents can only find, trust, and delegate to each other when capability descriptions follow one standard.
Discovery and routing
When a task arrives, orchestrators inspect published capability declarations to select the right specialist — no published card, no match.
Trust before invocation
Declared skills, provider identity, and cryptographic signatures reduce the uncertainty of calling a black-box AI service.
Multi-agent collaboration
In an A2A system, one agent reads another's card to decide whether to delegate a task. The metadata document serves as the handshake.
Developer Tools for Agent Metadata
Create, inspect, and understand the metadata that makes an A2A agent clear and discoverable.
Agent Card Generator
Create spec-compliant JSON from a guided form with endpoint, capabilities, skills, modes, and auth metadata.
Open toolAgent Card Validator
Paste JSON or check a live base URL for required fields, endpoint quality, auth clarity, and discovery readiness.
Open toolCode & JSON Examples
Start from practical templates for support, booking, shopping, coding, research, and internal workflow agents.
Open toolSchema Guide
Understand identity, supported interfaces, capabilities, default modes, skills, provider, and security fields.
Open toolWell-known URL Checker
Check both /.well-known/agent.json and /.well-known/agent-card.json for A2A discovery compatibility.
Open toolCore Capabilities of Agent Card & A2A
Built on the open Agent2Agent (A2A) protocol initiated by Google and governed by the Linux Foundation's Agentic AI Foundation (AAIF), enabling interoperable communication and collaboration across diverse frameworks.
Automated Capability Discovery
Standardized JSON discovery documents allow client systems and autonomous orchestrators to inspect interfaces, skills, and authentication boundaries before invoking a service.
Task Lifecycle Management
Structured state handling supports synchronous responses as well as asynchronous, long-running operations with streaming updates and push notifications.
Multi-Modal Communication
Declare default input and output media types across text, audio, video, and binary artifacts, with optional per-skill overrides for specialized models.
Multi-Agent Collaboration
Share execution context, structured data parts, and intermediate artifacts to enable complex multi-agent workflows without vendor lock-in.
Enterprise Security & Auth
Formal declarations for API keys, HTTP auth, OAuth 2.0 flows, and RFC 8785 JWS cryptographic signatures establish trust before calling any remote agent.
Content Mode Negotiation
Negotiate content modes and interactive formats between client and server agents, ensuring fluid, predictable user interface rendering.
Understanding the Agent2Agent (A2A) Protocol Architecture
The open A2A specification, governed under the Linux Foundation's Agentic AI Foundation (AAIF), standardizes communication between intelligent agents across frameworks. IBM's ACP has merged into A2A.
View Agent2Agent (A2A) Specification1
Client Agent Role - Client agents interpret user requests, match tasks against published skill declarations, and dispatch operations to remote endpoints using standardized bindings.
2
Remote Agent Execution - Remote agents receive tasks, execute domain logic across local models or external tools, and stream structured results back to the caller.
3
Built on Open Web Standards - Built entirely on established protocols including HTTPS, JSON-RPC, gRPC, and JSON, making integration straightforward across any technology stack.
Protocol Workflow with Agent Cards
- 1
Discovery & Inspection
Endpoints publish machine-readable metadata at well-known URIs, allowing callers to verify version compatibility and required credentials.
- 2
Asynchronous Task Flow
Task objects track the complete lifecycle from submission through intermediate progress updates to final artifact delivery.
- 3
Agent2Agent Collaboration & User Experience
Agents exchange structured messages and negotiate media formats in accordance with the specification, ensuring consistent multi-agent user experiences.
Who Uses A2A Metadata
From solo builders to compliance teams, the same document answers different questions.
AI developers
Define, debug, and distribute agents as reusable components with a spec-shaped discovery document.
Architects and DevOps
Deploy, configure, and monitor agent infrastructure from machine-readable metadata.
Security and compliance
Audit which agents hold which permissions and how data flows, straight from published metadata.
End users
Read the human-friendly view to see what an assistant can do and where its limits are.
Technical Guides
Technical Guides & Protocol Deep Dives
Practical engineering tutorials, protocol specifications, and security best practices for building production-ready A2A integrations.
Security11 min read
Signed Agent Cards: preventing poisoning and spoofing with JWS
Learn how A2A v1.0 uses JSON Web Signatures (JWS) in the signatures array to protect Agent Cards from tampering, poisoning, and impersonation, with a signing example that actually verifies.
Architecture10 min read
Unified agent architectures: A2A and MCP in practice under the AAIF
Discover how MCP and A2A now sit under the Linux Foundation's AAIF as a dual-layer architecture, pairing tool access with cross-framework agent discovery.
Ecosystem9 min read
A2A x402 extension: enabling pay-per-call monetisation for AI agents
Learn how the A2A x402 extension lets an agent declare payment support in its Agent Card and charge for skills with on-chain stablecoin payments, using task states and message metadata.
Frequently Asked Questions
Key technical details about the A2A protocol and agent discovery.
What is Agent Card in the Agent2Agent protocol?
+
Agent Card is a structured data format defined in JSON that describes the capabilities, features, and services of an AI agent. It is the foundation for capability discovery between agents in the Agent2Agent protocol, allowing client agents to find and utilize the most suitable remote agent for completing specific tasks.
How does the Agent2Agent protocol differ from other interoperability protocols?
+
The Agent2Agent (A2A) protocol and Agent Card technology complement protocols like MCP (Model Context Protocol). Both operate under the Linux Foundation's Agentic AI Foundation (AAIF) to form an open dual-layer infrastructure: MCP handles North-South interaction (connecting an agent downwards to tools, APIs, and data), while A2A and Agent Card handle East-West collaboration (connecting independent agents across frameworks and organizations).
Who is involved in developing the A2A protocol and discovery standards?
+
The Agent2Agent protocol and capability discovery specification were initiated by Google, contributed to the Linux Foundation in 2025, and joined the Linux Foundation's Agentic AI Foundation (AAIF) in 2026. Its Technical Steering Committee includes AWS, Cisco, Google, IBM Research, Microsoft, Salesforce, SAP, and ServiceNow, and IBM's Agent Communication Protocol (ACP) merged into A2A — an open, community-governed standard for agent interoperability.
What are the benefits of using standardized agent discovery cards?
+
Standardized capability declarations enable seamless collaboration between autonomous systems, increase automation reliability, and reduce cross-framework integration overhead.
What's the difference between Agent2Agent and MCP?
+
Under the Linux Foundation AAIF framework, A2A and MCP form the twin pillars of multi-agent architecture. MCP (Model Context Protocol) connects agents to internal tools, plugins, and data sources (North-South connectivity). A2A (Agent2Agent) uses Agent Cards, published at /.well-known/agent-card.json, to enable independent agents to discover each other, negotiate capabilities, authenticate, and collaborate across boundaries (East-West connectivity). They complement each other seamlessly.
Build and Publish with Confidence
AgentCard.net provides an integrated suite of generators, schema validators, code examples, specification guides, and well-known URL checkers.