Concept card

Multi-Agent Orchestration in Joule Studio

As of 2026-10-10

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

What is Multi-Agent Orchestration in Joule Studio?

SAP offers four documented ways to make agents work together: Joule Assistants that orchestrate SAP agents, subagents inside a Joule Studio project (max 5 direct, 25 total, sequential), agent requests from Joule capabilities including A2A calls to code-based agents, and A2A discovery in the new Joule Studio; n8n adds visual workflow orchestration.

Multi-agent orchestration means splitting a goal across several specialised agents and coordinating them: routing the request, passing context, collecting results and deciding when a human must step in. In SAP's stack in September 2026 there is no single "orchestrator product". There are four documented mechanisms, plus n8n for workflow-style orchestration, and each has hard limits you need to know before promising a design.

1. Joule Assistants orchestrating SAP agents

Since Sapphire 2026, SAP packages its delivered agents behind Joule Assistants. SAP SuccessFactors documentation spells out the mechanics: agents are assigned to assistants; when an assistant is enabled in SAP for Me, its agents become available without separate activation; the assistant "orchestrates the appropriate agents to fulfill business requests while respecting security and permission controls"; role-based permissions then decide which agents a given user can actually use. A documented example is the Job Change Request Agent, which interprets promotion and transfer requests and hands off to specialised agents when additional recommendations are needed. You configure this orchestration; you do not build it.

2. Subagents in Joule Studio, classic edition

In the agent builder you can add existing Joule agents as tools of another agent. SAP Help lists the restrictions:

A subagent cannot call back its parent (no cyclic dependencies).
Subagents run synchronously and sequentially, never in parallel.
An agent can have at most 5 direct subagents and 25 subagents in total, including nested ones.
Subagents must sit in the same Joule Studio project.

Why it matters

Documented hard limits (5 direct / 25 total subagents, sequential execution, 60-second synchronous A2A, 5-minute agent-request pause) decide which multi-agent designs are feasible on SAP today.
Joule Assistants make orchestration of SAP-delivered agents a configuration and permission task, not a build task.
A2A is SAP's supported bridge for pro-code and third-party agents, and it is one of the governed paths named in SAP's API policy.

Key points

Joule Assistants orchestrate SAP-delivered agents; enabling an assistant in SAP for Me makes its agents available, role permissions restrict use.
Classic subagents: no cycles, synchronous and sequential, max 5 direct and 25 total, same project only.
Content-based agents: YAML in Joule capabilities, max 100 KB per file, per-tool human approval, agent request pauses up to 5 minutes, no test traces.
Code-based agents: A2A 0.3.0 message/send, text only, 60-second synchronous limit, async push notifications need IAS App2App trust.
New Joule Studio agents expose an agent card, discover other agents over A2A, and reach SAP via MCP tools on the Agent Gateway.
Default to a single agent with a small toolset (3-5 tools); split into subagents or A2A peers only for a concrete isolation, ownership or reuse reason, not for architectural novelty.
n8n inside Joule Studio (GA announced for Q3 2026 by trade press, not confirmed in SAP sources as of 2026-10-10) is for known, fixed sequences with occasional agent steps — it removes handoff and timeout failure modes that pure agent-to-agent chains carry.
A subagent shares its parent's environment guardrail profile; an A2A peer is a separate trust boundary with its own Agent Card, auth scheme and audit trail — this maps to who should own each specialist, not just how it is called.

Terms used on this page

Subagent: A Joule agent added as a tool to another agent in the same Joule Studio classic project; runs synchronously and sequentially.
Content-based agent: Agent defined in YAML within a Joule capability and created in an agent runtime when the assistant is deployed.
Code-based agent (BYOA): Agent built with a third-party framework and called by Joule over the A2A protocol.
Agent card: A2A metadata document (name, description, version, URL, skills) that lets other agents discover and call an agent.
Joule Assistant: SAP-delivered orchestration layer that groups related agents (e.g. HR, finance) and exposes them together once enabled in SAP for Me, under role-based permissions.
n8n canvas (in Joule Studio): Managed, embedded visual workflow environment combining deterministic steps and agent calls, targeted for GA in Joule Studio in Q3 2026.
Supervisor-and-specialists pattern: Multi-agent design where one parent agent decomposes a goal and delegates narrow sub-tasks to specialist subagents or A2A peers, then assembles their results.
Fan-out: The number of subagents or tool branches a single orchestrating agent calls; SAP's 5-direct/25-total ceiling is the documented limit on fan-out in Joule Studio classic.

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 →

Guides that answer with this page

These guides cite this page as one of the sources their answer rests on.

Joule Studio SAP: availability, skills, guardrails