FerentinFerentin
Home/Glossary/A2A (Agent2Agent)

Glossary

What is A2A (Agent2Agent)?

An open protocol, now governed by the Linux Foundation, that lets AI agents built on different frameworks and run by different organizations discover each other and work together. An agent publishes an agent card describing its skills and endpoint, and other agents send it messages and track the tasks it runs. Where MCP connects an agent to tools and data, A2A connects an agent to other agents.

In depth

A2A (Agent2Agent) is an open protocol for communication between autonomous AI agents. Each agent publishes an agent card, a JSON document listing its name, skills, endpoint, supported protocol versions and how callers must authenticate. A client agent sends JSON-RPC messages to that endpoint, and the remote agent answers with a message or a task: a unit of long-running work with a lifecycle, status updates and artifacts such as files or structured results. Versions 0.3 and 1.0 are both in production use, with different method names on the wire. A2A is complementary to MCP: MCP lets an agent call a tool, while A2A lets an agent delegate work to another agent that has its own reasoning, tools and context.

Why it matters

A2A makes another organization’s agent part of your AI system. The remote agent decides for itself what to do with the data you send it, so the controls that matter for tool calls matter more here: who may call which agent and skill, what data may leave in the message, what comes back, whether the agent you approved is still the agent you are talking to, and an audit record of each exchange. Ferentin governs A2A 0.3 and 1.0 in both directions through the same policy, approval, data-loss-prevention and audit chain as MCP tool calls.

Common use cases

  • Delegating a task from an internal agent to a vendor’s specialist agent
  • Exposing an installed agent to external A2A clients through a governed gateway
  • Applying per-user skill permissions and approvals to agent-to-agent messages
  • Scanning messages and returned files for sensitive data in both directions
  • Recording each agent-to-agent call with the skill, protocol version and acting user

How Ferentin handles it

Ferentin is the trust layer for AI agents. The platform centralizes identity, policy enforcement and audit across LLMs, MCP servers and AI tools. A2A (Agent2Agent) is one of the primitives this trust layer is designed around. See the platform overview for how it fits into the service edge, control plane and observability plane.

Related terms

Agent Card

The JSON document an A2A agent publishes, conventionally at /.well-known/agent-card.json, describing its name, provider, skills, endpoint, supported protocol versions and authentication requirements. Clients read the card to decide whether and how to call the agent. Cards can be signed, so a client can check that the card came from the agent’s publisher and was not altered. Because a card can change after an agent is approved, including its endpoint and skills, a governed deployment pins the publisher’s signing keys, re-checks the card on a schedule and blocks the agent when its signature stops verifying or its endpoint moves.

MCP (Model Context Protocol)

An open protocol that standardizes how AI models connect to external tools, data sources and services. MCP defines a client-server architecture where AI applications (clients) communicate with tool providers (servers) through a structured interface, enabling AI agents to take actions in the real world.

MCP Gateway

A centralized control plane for Model Context Protocol traffic. An MCP gateway secures, manages and routes connections between AI agents and MCP servers, enforcing zero-trust access controls and providing visibility into every tool invocation.

Agentic AI

AI systems that can autonomously plan, decide and take actions to achieve goals, going beyond simple question-answer interactions. Agentic AI systems may invoke tools, access data sources, make API calls and chain multiple steps together, making governance and access control critical.

Data Loss Prevention (DLP) for AI

The detection, redaction or blocking of sensitive data, such as payment card numbers, bank account numbers, government identifiers and credentials, as it flows into and out of AI systems. DLP for AI applies to prompts and model responses, to the arguments an agent passes to tools and the results that come back, and to messages exchanged between agents.

Add the trust layer to your AI stack

Zero trust policy enforcement from prompt to tool call. Get started in minutes.

Get Started FreeBook a Demo