Concept card

A2A — Agent-to-Agent protocol in Joule

As of 2026-07-24T14:00:00Z

Edited by Cédric-Olivier Mâry, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is A2A — Agent-to-Agent protocol in Joule?

A2A lets a Joule orchestrator delegate a sub-task to an external agent — a third-party risk provider, Microsoft Copilot, Google Document AI — via a published Agent Card, while SAP data itself never leaves the BTP boundary; only instructions and results cross.

What it is

Agent-to-Agent (A2A) is an open, vendor-neutral protocol that lets one autonomous AI agent discover, invoke, and coordinate with another agent running on a completely different platform. Google announced it in 2025 and quickly donated it to the Linux Foundation, where governance now includes SAP alongside AWS, Cisco, Microsoft, Salesforce, and ServiceNow. Inside SAP, Joule adopted A2A support so that a Joule-hosted orchestrator agent can hand off a sub-task to an agent that SAP does not own and cannot see inside — a supplier-risk agent from a third-party provider, a Microsoft Copilot approval agent, or a Google Document AI extraction agent.

The cleanest mental model: Model Context Protocol (MCP) is how a single agent reaches into its own toolbox of databases, APIs, and files it is allowed to touch. A2A is how that same agent picks up the phone and asks a different agent, living on a different runtime with different owners, to do a piece of work it cannot do itself. MCP is the hand; A2A is the voice.

Why it matters

Why it matters in practice

SAP is a founding Linux Foundation member alongside AWS, Cisco, Microsoft, Salesforce and ServiceNow — a genuinely cross-vendor protocol, not SAP-only, in Joule from Q1 2026.
The MCP-vs-A2A distinction is precise and useful: MCP is an agent's hands (tools/data), A2A is its voice to other agents — a clean way to explain the architecture to a client.
The four-step protocol flow (Agent Card fetch, Task Request, Task Response, merge) supports both polling and webhook callbacks — relevant for long-running document or multi-party approval workflows.

Key points

A2A — Agent-to-Agent protocol in Joule is mastered only when it changes a named buyer decision.
Start with the semantic contract and control model before demonstrating the tool.
Use current SAP, analyst, study, KG, and news signals as evidence, not decoration.
Separate verified facts from directional trends and modeled assumptions.
Define owner, metric, threshold, support path, and rollback before scaling.
For AI use cases, measure reliability, cost, latency, safety, and human validation.
Leave a reusable operating asset: memo, checklist, control table, and exception log.
A premium answer is short, trade-off explicit, and defensible in a steering committee.

Terms used on this page

Decision owner: The accountable person who accepts the trade-off and funds the next action.
Semantic contract: The shared definition of business terms, metrics, entities, and access rules used by tools and teams.
Control plane: The layer that applies policy, access, lineage, monitoring, and escalation across the operating model.
Evidence grade: A label that separates verified fact, directional signal, modeled assumption, and field observation.
Adoption metric: The measurable behavior proving that the concept changed actual work after go-live.
Agent reliability: The consistency, cost, safety, and policy compliance of an agent across repeated runs.
Decision owner: The accountable person who accepts the trade-off and funds the next action.
Semantic contract: The shared definition of business terms, metrics, entities, and access rules used by tools and teams.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks · the facts worth quoting.

Open in the app →