Flagship Workflow: An Agent System That Changes Over Time

This is the flagship use case for yggdrasil.

Scenario

A support workflow starts simple:

  • one intake agent
  • one billing playbook
  • one normalization tool

Later, policy changes:

  • high-value billing disputes must escalate to manager review
  • the playbook changes
  • the routing policy changes

The application should not need to be conceptually rewritten. The graph should change, and the runtime should be able to explain the result.

Why yggdrasil Fits

This workflow combines:

  • graph-defined composition
  • workflow state
  • explainable context selection
  • deterministic routing
  • graph evolution over time

Runnable Example

See:

That example demonstrates a multi-agent pipeline where an intake agent hands off work to a specialist, showing graph-defined composition and explainable execution.

What This Shows

The important value is not only that the workflow runs.

The important value is that you can answer:

  • what changed?
  • why did the system route differently?
  • which context was active in version 1 vs version 2?
  • which policy change caused the new path?

That is the control-plane story.


Copyright © 2026 Hoang Dao. MIT License.

This site uses Just the Docs, a documentation theme for Jekyll.