Discover, don’t describe.
Tracey captures your computational workflows from exploration through experimentation to publication. No annotations, no workflow languages, no changes to how you work.
The workflow that gets shared is rarely the one that actually ran.
The most common cause of failed reproducibility isn’t missing data or incorrect code. It’s a process that was never recorded. Researchers run ad hoc commands, revise scripts between executions, and pursue approaches they later discard. By the time results are ready to share, the workflow gets reconstructed largely from memory. Existing tools help at the edges, but the critical middle, where research actually happens, goes unrecorded.
No new habits required.
Always-on capture
Runs as a background daemon and records every execution automatically, including the exploratory dead ends.
HPC and Slurm ready
Native Slurm integration captures batch jobs on compute nodes without modifying your job scripts.
Portable artifacts
Derives a workflow DAG from your execution history and packages everything into a self-contained Apptainer container.
Built for how research actually happens.
Most tools ask you to describe your workflow before you run it, or to package it after the fact. Tracey does neither. It observes real executions and derives workflow structure from them, so the pipeline you share reflects what you actually did, not what you remember doing.
Ready to try Tracey?
Tracey is developed at the Research Computing Center at the University of Chicago. Access is currently available on request.