On a visual canvas, you drag‑and‑drop nodes for inputs/outputs, AI agents, human reviews, decisions, data connectors, REST calls, and custom Python to create your workflows.  

Workflows can then be run as Jobs, which execute with logs, metrics, and tracing so your automations remain understandable, testable, and auditable.

---

✨ The Challenge

Design a reusable “Intake → Understand → Decide → Review → Deliver” automation that tackles the messy reality of multi‑format intake, mixed rule‑based + judgment decisions, and exception handling—while keeping every step traceable. Public/demo data and free external APIs are fine; enterprise integrations are not expected.

Want to stand out with your submission? 

Check out the detailed challenge guidelines here.

---

🛠️ What You’ll Build

Create one end‑to‑end Opus workflow that tackles a real use case in an industry of your choice. Demonstrate it with realistic sample inputs.

Required building blocks:

• • Data import & processing: Import from an external source (file, sheet, or public API) and make processed records available downstream.
  

• 
  

• • Decisioning & routing: Combine deterministic rules (thresholds, required fields, format checks) with AI reasoning (categorization, summarization). Support multi‑condition logic, parallelize independent steps, and handle errors/timeouts sensibly.
  

• 
  

• • Review for quality & safety: Include at least two checkpoints—Agentic Review (policy/guidance checks) and Human Review for low‑confidence or high‑impact cases.
  

• 
  

• • Provenance & audit: Generate a compact audit artifact (JSON or PDF) capturing inputs, extracted fields + confidence, rules fired, scores/rationales, review actions, timestamps, IDs, and any external source URLs.
  

• • No‑code: prompt‑to‑workflow generation and the visual canvas builder.

• • Developers: Opus Code node (custom Python), External Service node (public APIs), and Data Import/Export for supported platforms; jobs can be triggered via API.