taskflow

Showcase

End-to-end case studies — real problems solved with taskflow, built up phase by phase.

The Showcase is where the pieces come together. Each case study starts from a concrete problem you have probably already hit, builds a taskflow one phase at a time, and explains why each phase looks the way it does. By the end of each one you have a complete, saveable flow you can run yourself.

They are not host walkthroughs — the same flow runs identically on Pi (/tf run) and on Codex / Claude Code / OpenCode / Grok Build (taskflow_run). They are pattern guides: read them to learn the shape, then adapt the shape to your own problem.

New here? The What is taskflow? page explains the model in one page, and Getting Started gets a flow running in five minutes. Come back here when you want to see the model stretched across a real problem.

The case studies

Three problems, three shapes. Each one motivates a different part of the runtime — fan-out, tournament selection, and dynamic planning with loops.

What you will take away

Each case study is structured the same way so you can map the lessons onto your own work:

  1. The problem. Why the obvious approach (one big agent call, or three manual delegations) does not scale.
  2. Phase by phase. Each new phase is added with a reason — fan-out for throughput, a gate for safety, a loop for convergence. You see the flow grow, not just the finished artifact.
  3. The complete flow. The assembled JSON, ready to save and run.

Want the short version of why a declarative graph beats an imperative script — with concrete before/after numbers? See Why taskflow?.

Next steps

Last updated on

Was this helpful?

Help us improve the docs or ask a question in the community.

On this page