Agentic AI on BTP: Chaining Agents into a Workflow
Third in the author's BTP agent series: it replaces one orchestrator agent that decided its own path, which under an unattended schedule skipped messages, re-drafted replies already sent and carried more tool access than a step needed, with a workflow, a declared main line of steps and branches that is saved once and runs without a model reordering it. A step can be an agent with its own tools and model, or one of four deterministic kinds, condition, transform, http and python, that run without one, so judgement and plain logic share the same line; a fan-out step with per-item branches is what turns one workflow into the author's mail-triage pipeline. That is the agent-toolingAgent toolingBuilding an agent into your own application — the tools it may call, the state it keeps between turns, and the framework it runs in. Distinct from MCP servers, which expose an application to somebody else's agent, and from agent guidance, which is knowledge handed to a coding agent.SAP and community, unassessed case for a framework holding state between turns: before a run starts every agent and peer the workflow can reach is checked for a usable credential, and the run itself is recorded per item as succeeded, failed, skipped or partial rather than one status.
Where it sits
Who is behind it
In the atlas: maintains 25, contributes to 39, wrote 18@lemaiwoauthor
Properties
About this work
Details
Graph
Source
History
- Entered the atlasDiscovered via SAP Community · agentic-ai-tag, SAP Community · btp-tag.
For your agent
- Ref
content:agentic-ai-on-btp-chaining-agents-into-a-workflow- JSON
- /content/agentic-ai-on-btp-chaining-agents-into-a-workflow.json
- MCP
- Connecting an agent