Dyno
Dyno Phi

Filter and rank
your protein designs.

Agentic skills for calibration and advancement of AI-generated protein designs, without managing infrastructure.

01

Design

Upload binder candidates generated by any tool — RFDiffusion, BoltzGen, or your own model — via the web interface, CLI, or Claude Code.

02

Score

Run scoring models on Dyno cloud GPUs to evaluate each design.

03

Filter

Shortlist and rank candidates against experimentally calibrated filters.

04

Advance

Move shortlisted designs to synthesis, hand off to the wet lab, or share with collaborators.

Access

Two ways to use Dyno Phi — pick the interface that fits your workflow. Both submit jobs to the same filtering pipeline and share the same datasets.

Quickstart

  1. 1
    Sign up — Create an account at design.dynotx.com.
  2. 2
    Create an API key — Go to Settings in the dashboard and generate a key.
  3. 3
    Install the CLI
    # Requires Python ≥ 3.9
uv tool install dyno-phi

    New to uv? docs.astral.sh/uv — use uv tool install to make phi available globally without activating a virtualenv.

  4. 4
    Set your API key
    # Get key at design.dynotx.com → Settings
export DYNO_API_KEY="your_key_here"
  5. 5
    Run the tutorial
    # Example files and tutorial
phi tutorial

    Downloads example datasets and walks you through the full filtering pipeline with real data.

  6. 6
    Filter your own candidates
    # Upload binder PDB / FASTA files
phi upload ./designs/

# Run filter pipeline
phi filter --preset default --wait

# View ranked scores
phi scores

    State is cached in .phi-state.json — after phi upload, subsequent commands pick up the active dataset automatically. Full CLI reference →