A collection of reproducible demos for Crewplane, a workflow runner for coordinating coding-agent CLIs.
Each lab is self-contained: it includes the workflow definition, the generated responses, and the execution record needed to inspect what ran. Use the labs to explore orchestration patterns, compare provider configurations, or adapt a working workflow for your own experiments.
| Lab | What it demonstrates |
|---|---|
| LeetCode 3348: eight-agent model and reasoning comparison | Runs the same algorithm problem across eight Codex model/reasoning configurations in parallel, then evaluates correctness, runtime, and token usage in a comparison stage. |
The included results are snapshots of individual runs, not general-purpose model benchmarks. See each lab's README for its methodology, artifacts, and caveats.
Find these examples useful? ⭐ Star Crewplane on GitHub to support the project.
You do not need Crewplane or a provider account to inspect the recorded runs. Inside each lab, start with:
README.mdfor the experiment overview and direct artifact links.crewplane/workflows/for the Markdown workflow definition.crewplane/execution-results/for consolidated node outputs.crewplane/execution-stages/for manifests, telemetry, provider logs, and per-agent responses
Follow the Crewplane README
to install Crewplane and set up a project. Then copy the workflow from the lab
you want to reproduce into your project's .crewplane/workflows/ directory,
update .crewplane/config.yml for the required providers and models, and run
the workflow.
Provider and model availability may differ by account. Results, latency, caching, and token usage will also vary between runs.
This repository is licensed under the MIT License.