Fully Automated Mode
Two steps. You configure the campaign here, run one command on a machine with a GPU, and bring back a single file to explore. Everything between those points — writing the spec, generating the YAMLs, preflighting, predicting, and analysing — is handled for you.
Prepare
Configure a run and download the bundle.
Describe your proteins, partners and ligands, choose how hard to push the hardware, and get back one self-extracting file. It carries the campaign spec, a pinned environment specification covering Boltz-2 and the whole analysis stack, and the scripts that run them. Nothing else to install and no versions to match.
Ends with a file on your machine.
Step 2Analysis
Upload the results file and explore.
The bundle writes exactly one results file when it finishes. Upload it and get a sortable table of every target, a confidence-versus-affinity triage plot, and per-target pose viewers with the protein-ligand interactions that were detected.
Starts from the file the bundle wrote.
What happens on your machine
The bundle is a single self-extracting file. Double-click it (or run it from a terminal) and it
installs a pinned environment via pixi,
then runs the campaign end to end and writes one .bmz results file next to itself.
The Boltz-2 model weights are downloaded on first use rather than shipped in the bundle —
they are large, versioned by Boltz itself, and cached in your home directory afterwards, so a
second campaign skips that step.