Technology · Agentic → Agent Protocols · Open Source · wiki:deep

Agent2Agent Protocol

Taxonomy

Agent2Agent (A2A) is an open protocol for communication and interoperability between opaque agentic applications — agents built on different frameworks, by different vendors, on separate servers — so they can collaborate as peers, not merely as tools. Agents discover capabilities via Agent Cards, negotiate modalities (text, forms, media), and collaborate on long-running tasks without exposing internal state, memory, or tools. Wire format: JSON-RPC 2.0 over HTTP(S), with sync, SSE streaming, and async push. Linux Foundation project (Google-contributed); Apache-2.0; SDKs in Python, Go, JS, Java, .NET, Rust.

RepoDocsStack: agentic-agent-protocols

Why it matters here

An agentic OS that only mounts tools still needs a peer handoff language when work crosses products or companies. A2A is the research pick for agent↔agent discovery and task exchange. Prefer MCP for host↔tool/resources/prompts; use both when specialists are external agents and tools are MCP servers. Protocol alone does not replace policy gates or the ledger.

How it works

Publish an Agent Card; peers discover and open authenticated sessions; clients send tasks/messages and receive artifacts via request/response, streaming, or push. Opacity: collaborate without sharing proprietary internals.

  1. Expose agents as A2A servers (ADK, LangGraph, BeeAI, …).
  2. Publish Agent Cards (capabilities + connection).
  3. Clients negotiate and run tasks (sync / SSE / push).
  4. Audit handoffs in the OS ledger; still gate world-binding actions.

Related: model-context-protocol · topics/33-skill-registry

Flow

Scroll inside the canvas to pan

When to reach for it

  • Use when: you need cross-stack agent collaboration without a shared runtime.
  • Skip when: you only need tool mounting (model-context-protocol) or in-process multi-agent frameworks.
  • Prefer instead: MCP for tools/resources; A2A when the peer is another opaque agent.

Limits

  • Interop ≠ authority: discovered skills can still be unauthorized — IAM/policy still apply.
  • Opacity cuts both ways: harder to debug peer internals; need contract tests.
  • Spec churn: track a2a-protocol.org version and SDK alignment.
  • Not the ledger: handoff records must still land in company SoR.

Linked from

Catalog backlinks — what points here (wiki “what links here”).

Features

Stacks

Peers

What we checked

Claims below are backed by science sources on disk.

Protocol / interoperability role

a2aproject/A2A repository · MODERATE

“An open protocol enabling communication and interoperability between opaque agentic applications.”

Spec notes

Agent2Agent (A2A) Protocol Specification · MODERATE

“The Agent2Agent (A2A) Protocol is an open standard designed to facilitate communication and interoperability between independent, potentially opaque AI agent systems.”

MCP vs A2A contrast

MCP vs A2A — coordination complexity · MODERATE

“A task represents a stateful and long-running piece of work whose progress needs to be tracked by the client and the server.”

In this research stack

Features and peers linked from the catalog map — not a second product surface.

Features it supports

Same cell

Stack

Document management

No DM vendor crosswalk edges yet.

Source

Primary repo github.com/a2aproject/A2A · Open Source

technologies/agent2agent-protocol/README.md

Research inventory

9 tags · 6 out · 7 in · 3 artifacts · 1 gaps · 26 corpus docs

Catalog tags

landscape.layer
Agentic
landscape.subcategory
Agent Protocols
license_tag
Open Source
maps.dm
absent
maps.features
5
one_liner
Agentic
review.depth
science
slug
agent2agent-protocol
title
Agent2Agent Protocol

Artifacts

  • dm_map · absent
  • features_map · present · technologies/agent2agent-protocol/features.md
  • readme · present · technologies/agent2agent-protocol/README.md

Out · alternative_to

Out · maps_to

In · alternative_to

In · in_stack

In · maps_to

Corpus tags

category
Agentic → Agent Protocols
dedication
open-source
feature
agent-identity-interoperability
apis-sdks-mcp-extensibility
mcp-tools
multi-agent-collaboration
r2-02-assistant-expert
kind
map_edge
tech_features
tech_quote
tech_readme
tech_science_source
tech_section
needs_deepen
false
quality
ok
section
Evidence
Features map
GitHub map
How it works
Limits & failure modes
Links
Scientific notes
What it is
When to use / skip
Why it matters here
slug
agent2agent-protocol
source_id
a2a-repo
a2a-spec
mcp-vs-a2a-2607-23884
technology
agent2agent-protocol

Corpus documents (26)

map_edge · 5

  • agent2agent-protocol → agent-identity-interoperability
  • agent2agent-protocol → apis-sdks-mcp-extensibility
  • agent2agent-protocol → mcp-tools
  • agent2agent-protocol → multi-agent-collaboration
  • agent2agent-protocol → r2-02-assistant-expert

tech_features · 1

  • Agent2Agent Protocol · features

tech_quote · 6

  • Agent2Agent Protocol · a2a-repo
  • Agent2Agent Protocol · a2a-spec
  • Agent2Agent Protocol · a2a-spec
  • Agent2Agent Protocol · a2a-spec
  • Agent2Agent Protocol · mcp-vs-a2a-2607-23884
  • Agent2Agent Protocol · mcp-vs-a2a-2607-23884

tech_readme · 1

  • Agent2Agent Protocol

tech_science_source · 3

  • Agent2Agent Protocol · a2a-repo
  • Agent2Agent Protocol · a2a-spec
  • Agent2Agent Protocol · mcp-vs-a2a-2607-23884

tech_section · 10

  • Agent2Agent Protocol · Evidence
  • Agent2Agent Protocol · Features map
  • Agent2Agent Protocol · GitHub map
  • Agent2Agent Protocol · How it works
  • Agent2Agent Protocol · Limits & failure modes
  • Agent2Agent Protocol · Links
  • Agent2Agent Protocol · Scientific notes
  • Agent2Agent Protocol · What it is
  • … +2 more