Skip to content
Get packs

Guide · Agent skill

How to use the Rowan molecular modeling agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The Rowan agent skill teaches an AI agent to run molecular modeling and medicinal-chemistry jobs on Rowan's hosted cloud service through its Python API. This guide is for chemists and developers who have never used agent skills: it explains what the skill does, how to install it, how to ask for work in plain language, and what to check before running anything. By the end you will be able to set up the skill, confirm your account works, and request jobs such as pKa prediction, conformer ensembles or docking.

Rowan Molecular ModelingK-Dense-AI/scientific-agent-skills · MIT

Quick start

  1. 1Download the ZIP. Download the Rowan skill ZIP from the directory page.
  2. 2Install the skill. Unzip it into ~/.claude/skills/ for all projects or .claude/skills/ for one project, or upload the ZIP under Settings, Capabilities, Skills in the Claude apps.
  3. 3Install the Python package. Run uv pip install with the pinned rowan-python version, and do not install the unrelated rowan package.
  4. 4Set your API key. Export ROWAN_API_KEY in your environment, then run rowan.whoami() to confirm the account and credit balance.
  5. 5Make a natural request. Describe the task, the input molecules and the output you want, and let the agent choose the workflow.
  6. 6Check the result. Review the returned structured data and confirm the input format and stereochemistry were correct.

What the Rowan skill makes the agent do

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches it. The Rowan skill covers pKa and macropKa prediction, conformer and tautomer ensembles, single-ligand and analogue docking, protein-ligand cofolding, MSA generation, molecular dynamics, membrane permeability, and molecular descriptor workflows.

Once installed, the agent writes and runs Python code that sends these jobs to Rowan's hosted service rather than your own computers. Results come back as structured data that the agent can read, summarise or feed into the next step of a pipeline.

  • Batch property prediction such as pKa, descriptors, permeability and solubility
  • Conformer and tautomer ensemble generation
  • Docking, including analogue series and pose refinement
  • Multi-step pipelines, for example tautomer search, then docking, then pose analysis
  • The skill also states when RDKit alone is enough, such as for simple file conversion

Install the Rowan skill from the ZIP

Download the Rowan skill ZIP from the directory page, then unzip it into one of two locations. Use ~/.claude/skills/ to make the skill available to all your projects, or a project's .claude/skills/ folder to limit it to one project.

If you use the Claude apps instead of a folder on disk, upload the ZIP under Settings, then Capabilities, then Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into .claude/skills/ inside the project folder
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Set up Python and your Rowan API key

The skill needs Python 3.12 or newer, the rowan-python package with its RDKit and stjames dependencies, a Rowan account with available credits, and network access. Install the package with uv, and lock the versions in your own environment so a campaign stays reproducible. Do not install the unrelated package named rowan.

Set your API key as an environment variable. The SDK reads it automatically. Then check that the account works and see how many credits remain.

  • Requires Python 3.12 or newer
  • Requires network access to Rowan's hosted service
  • Requires a ROWAN_API_KEY from an account with credits
uv pip install "rowan-python==3.2.0" export ROWAN_API_KEY="your_api_key_here" import rowan user = rowan.whoami() print(f"User: {user.email}") print(f"Credits available: {user.credits_available_string()}")

Ask for work in natural language

You do not need to write code yourself. Describe the chemistry task and the input you have, and name the workflow if you know it. The agent should pick the matching Rowan workflow and check the input format before submitting.

Give the agent SMILES strings or 3D structures, and say what you want back. Mention stereochemistry, charge and protonation state if they matter, because the skill notes that canonicalization alone does not settle these choices.

  • Predict pKa for these SMILES using a pKa method that accepts SMILES, and return a table
  • Generate a conformer and tautomer ensemble for this 3D structure, then dock the ensemble into my receptor
  • Compute MW, SLogP and TopoPSA for this list of molecules and flag any above my cutoffs
  • Run a batch docking campaign for these analogues and summarise the scores

Know which input format each workflow needs

The most common error is passing the wrong input type. SMILES strings suit topology-based methods such as some pKa methods, ADMET, LogP, macropKa and solubility. Geometry-based workflows need a 3D molecule with coordinates, so a bare RDKit molecule made from SMILES is not enough. Generate coordinates explicitly with rowan.Molecule.from_smiles(), or import an existing geometry.

Pose-analysis molecular dynamics needs both ligand SMILES and a protein complex that contains the bound pose. SMARTS is a substructure-query language and is not a workflow input, so convert InChI with a chemistry toolkit first.

  • pKa: some methods take SMILES, others need coordinates; the default method needs coordinates
  • Descriptors, tautomers, docking and analogue docking need a molecule with a conformer
  • Use RDKit to validate SMILES before submitting
from rdkit import Chem smiles = "CCO" mol = Chem.MolFromSmiles(smiles) if mol is None: raise ValueError(f"Invalid SMILES: {smiles}")

Run a first job and read the result

Every Rowan task follows the same pattern: submit a workflow, wait for it to finish, then read typed results. The example below calculates descriptors for aspirin. It consumes credits, so run it on an account where you are willing to spend them.

For long jobs, streaming polls for snapshots while the job runs. Streaming is polling, not a server push, and partial results may be incomplete, so check the complete flag before using them.

  • result() blocks until the job is done
  • stream_result() gives snapshots; use .data until .complete is true
  • Failed and stopped runs also count as done
import rowan mol = rowan.Molecule.from_smiles("CC(=O)Oc1ccccc1C(=O)O") wf = rowan.submit_descriptors_workflow(mol, name="aspirin") result = wf.result() print(result.descriptors["MW"]) print(result.descriptors["SLogP"]) print(result.descriptors["TopoPSA"])

Safety notes: scripts, network and credits

This skill runs scripts and uses the network. Every hosted job is sent to Rowan's service and consumes credits from your account. Review the code the agent proposes before it runs, and check your credit balance before large batches.

Licensing needs care. The repository is listed under MIT, but the skill's own metadata marks the Rowan service as proprietary and requires an API key. Check the terms that apply to your use before running work for a company or publishing results.

  • Do not use the skill for offline or air-gapped work; it needs network access
  • Keep your ROWAN_API_KEY in an environment variable, not in code you share
  • Lock the rowan-python version for reproducible campaigns

What people use it for

  • Predicting pKa for a library of SMILES before choosing which compounds to synthesise
  • Generating tautomer and conformer ensembles, then docking them into a prepared receptor
  • Calculating descriptors such as MW, SLogP and TopoPSA for batch filtering
  • Running a multi-step pipeline from tautomer search to docking to pose analysis
  • Analogue triage across a small series of related molecules, with workflow IDs recorded for later review

Frequently asked questions

Is the Rowan agent skill free to use?

Rowan's published terms list a free tier with signup credits and a weekly allowance, but the free tier does not include every function. Advanced workflows such as macropKa, MSA, membrane permeability, protein MD and batch docking are listed as advanced access, and some offerings are enterprise-only. Check your own account plan with rowan.whoami() because allowances differ for academic and team accounts.

What do I need to run the Rowan skill?

You need Python 3.12 or newer, the rowan-python package with its RDKit and stjames dependencies, a Rowan API key from an account with available credits, and network access. The skill cannot run offline because the workflows execute on Rowan's hosted service.

How do I install the Rowan skill in Claude?

Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

Can I use Rowan instead of RDKit?

Not for everything. The skill says RDKit alone is enough for simple molecular file conversion and basic structure handling. Use Rowan when you need hosted pKa, docking, conformer ensembles, molecular dynamics or other modeling workflows that RDKit does not provide.

Why does my Rowan job fail with a validation error?

The most common cause is the wrong input type, such as a bare SMILES string passed to a workflow that needs a 3D molecule. Generate coordinates with rowan.Molecule.from_smiles() or import a geometry, and check the troubleshooting reference included with the skill. An HTTP 422 response is a request-validation failure, so fix the input rather than assuming the feature is disabled.

Is the Rowan skill safe to use with confidential molecules?

The skill runs scripts and sends your molecular inputs to Rowan's hosted service over the network. Before using confidential structures, read Rowan's terms and confirm they fit your data-handling rules. Keep your API key in an environment variable and review the agent's code before it runs.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

The Rowan skill suits medicinal or computational chemists who screen many molecules, and developers building multi-step pipelines that need hosted compute instead of local HPC or GPU setups. It is a poor fit for simple file conversion where RDKit is enough, for offline or air-gapped work, and for anyone without a Rowan API key and credits.

Download the skill

Find exactly what
you're looking for