Skip to main content

Cognitive Chemistry Labs

Focus on the science. Let us handle the rest.

Ready-made workflows and an Assistant that explains every step. Start at $0 and pay as you go for the compute you use.

Describe your problemBrowse workflows
Or schedule a demo

  • 15

    Ready-made workflows

  • 4

    Problem families

  • 63

    Tools

  • 10

    CPU & GPU types

[Virtual screening]Fig. 01
  • Screen a compound library against a targetSMILES → 3D embed → dock
  • Find top binding poses with diffusion dockingSMILES → DiffDock → top poses
  • Screen an ultra-large combinatorial libraryThompson Sampling → 3D embed → dock
  • Rescore docking hits with a CNN scoring functionVina triage → GNINA rescore
  • Re-predict docking hits with Boltz-2 affinitiesVina triage → Boltz-2 re-predict (batch)
  • Re-dock top hits with surface-aware diffusionVina triage → SurfDock (scatter)
Workflows
6
Stages
5
Longest pipeline
6 steps
Find the compounds in a library most likely to bind your target, then rescore the best hits with a stronger method. Each line is a ready-made workflow; lines share a station where they run the same step. See the virtual screening workflows

Workflows

Start from the problem, not the plumbing.

Each workflow is a tested pipeline, named for the problem it solves. Bring your inputs: the steps, file formats, and hand-offs between tools are already connected.

Family 01 · 6 workflows

Virtual screening

Find the compounds in a library most likely to bind your target, then rescore the best hits with a stronger method.

Explore virtual screening

  • Screen a compound library against a target

    SMILES → 3D embed → dock

    3 steps

  • Find top binding poses with diffusion docking

    SMILES → DiffDock → top poses

    2 steps

  • Screen an ultra-large combinatorial library

    Thompson Sampling → 3D embed → dock

    4 steps

  • Rescore docking hits with a CNN scoring function

    Vina triage → GNINA rescore

    5 steps

  • Re-predict docking hits with Boltz-2 affinities

    Vina triage → Boltz-2 re-predict (batch)

    5 steps

  • Re-dock top hits with surface-aware diffusion

    Vina triage → SurfDock (scatter)

    6 steps

Mapped in Fig. 01 above.

Family 02 · 4 workflows

Structure prediction to docking

Dock against a protein with no solved structure: predict it from sequence, find its pockets, then dock.

Explore structure prediction to docking
[Structure prediction to docking]Fig. 02
  • Dock into a predicted structure without choosing a boxSequence → fold → DiffDock
  • Dock against a protein with no solved structureSequence → fold → dock
  • Dock into a cleaned predicted structure and see the contactsSequence → fold → prepare → dock → PLIP
  • Compare docking against a wild type and its point mutantsPoint mutants vs wild type → fold → dock → compare

Family 03 · 4 workflows

Library design

Generate or enumerate new compounds for a target or an SAR series, then dock them.

Explore library design
[Library design]Fig. 03
  • Dock analogs enumerated from your SAR dataFree Wilson → 3D embed → dock
  • Grow analogs from your SAR table with matched pairs, then dockMatched pairs → analogs → dock
  • Generate and dock new ligands for a pocketDrugFlow generate → dock
  • Build a library from building blocks, filter it, then dockReaction enumerate → filter → diversity pick → dock

Family 04 · 1 workflow

Predictive modeling

Pull measured activities for your targets, curate them into a clean dataset, and train a model that predicts activity for your own compounds.

Explore predictive modeling
[Predictive modeling]Fig. 04
  • Train an activity model from ChEMBL data for your targetsChEMBL activities → curate → train MPNN

  • Build your own

    Open any template in the visual builder and change its steps, or start from a blank canvas. The builder checks each connection as you make it.

    Open the builder

  • Or let the Assistant build it

    No template fits? Describe the problem, and the Assistant chains the steps itself, checks the connections, and prices the plan. It runs only when you approve.

    Describe your problem

A different model

Computational chemistry, without the seat license.

Commercial modeling suites are typically sold as annual per-seat licenses. CogniChem runs in your browser and bills only the compute you use.

See plans and rates
Cost to startThe old way · An annual seat license, often priced only by quoteCogniChem · $0. A plan is optional: it lowers your rates.
Getting startedThe old way · Procurement, a sales call, then an install on every workstationCogniChem · Sign up and run in your browser today
HardwareThe old way · Your own workstations or cluster, GPUs includedCogniChem · CPU to Nvidia B200, on demand, chosen per step
What you pay forThe old way · Seats, whether or not anyone runs anythingCogniChem · The compute you use, from your wallet
PipelinesThe old way · Wiring tools and file formats together yourselfCogniChem · 15 ready-made workflows, every step connected and tested
Learning the methodsThe old way · Training courses and manualsCogniChem · An Assistant that explains each step and cites the method papers

What it costs

A run costs cents to dollars, not a contract.

What a typical run reserves on the free Basic plan: the catalog's expected runtime × the per-second rate. You're charged only for the seconds it actually runs, and higher plans pay less.

All compute rates

Bigger jobs: an Nvidia H100 is $13.82 an hour on Basic and $7.88 on Pro, billed by the second.

CogniChem Assistant

Ask it why.

Tell the Assistant what you want to find out. It picks a ready-made workflow or builds one for the job, explains why each step is there, points you to the papers behind each method, and prices the plan. Nothing runs until you click Approve.

  • Answers with sources

    Ask why a step is there and the answer comes from the papers behind the method. Every compound or target it looks up links to the source record.

  • It proposes; you approve

    The Assistant can plan, estimate, and propose, but it cannot start, cancel, or delete anything. Your Approve click starts the run, re-priced at that moment.

  • A spend cap on every conversation

    $1.00 of Assistant tokens by default. Only you can raise it, $1.00 at a time.

  • Your files and live data

    Attach PDB, SDF, or SMILES files, and look up compounds and targets in PubChem, ChEMBL, Open Targets, UniProt, and the PDB.

Describe your problemHow the Assistant works
[assistant exchange]Fig. 05
Why does this workflow dock everything with Vina, and only then run Boltz-2?

Vina is fast and runs on CPUs (about $0.13 a typical run), so it can screen the whole library. Boltz-2 predicts the full protein–ligand complex and a binding affinity, which tells you more but needs a GPU (about $1.93 a run on an Nvidia A10). So the workflow spends Boltz-2 only on the compounds Vina ranks highest.

Proposed workflow · waiting for you

Re-predict docking hits with Boltz-2 affinities

Vina triage → Boltz-2 re-predict (batch)

Steps
5
Runs on
CPU · A10 GPU
Estimate
Priced for your plan
Illustration of an exchange. Prices and papers are the catalog's. In the app, the plan card shows the estimate the API computed for your plan, and approving re-checks it.

How it works

Three steps from question to results.

  1. Pick a workflow, or describe the problem

    Browse workflows by the problem they solve, build your own in the visual builder, or tell the Assistant what you need and let it pick or build one.

  2. Add your inputs and check the estimate

    Upload a structure and a compound list, or reuse files from earlier runs. The estimate for your plan appears before you launch, and the run pauses at its cost cap instead of overspending.

  3. Run it and collect the results

    Follow each step as it runs. Results are saved as files you can download or feed into your next workflow.

Why CogniChem

You pay for results, not for wiring.

The hard part of computational chemistry is rarely one model. It is getting formats, conversions, and hand-offs right between five of them. That part is done before you arrive.

Working pipelines, not parts

Workflows ship with every step connected and tested, so the first run is a result, not a debugging session.

No infrastructure to run

Jobs run on on-demand CPUs and GPUs, billed by the second at your plan's rate. Nothing to install or keep running.

Priced up front

Every job and workflow is estimated before it starts, and workflow runs pause at their cost cap.

Built for individuals and teams

Individual researchers, academic groups, and small to mid-sized companies get these tools without enterprise pricing. Plans start free.

Pricing

Start free. Pay as you go for compute.

Plans set your rates, monthly credit, and limits. Compute is billed per second from your wallet, at lower rates on higher plans.

Compare plans

  • Basic

    Starting at $0

    Get started with basic features.

    H100 GPU · $13.82 / hour

  • Starter

    $100 / month per seat

    For small teams needing more power.

    H100 GPU · $9.86 / hour

  • Pro

    $500 / month per seat

    Advanced features for research groups and startups.

    H100 GPU · $7.88 / hour

  • Enterprise

    Contact us for pricing

    Custom solutions for large organizations and commercial use.

    H100 GPU · $5.94 / hour

For experts

63 tools when you want one method.

Every model behind the workflows runs on its own too, along with quantum chemistry, thermodynamics, and custom model training.

21 tools · Modeling and analysis tools for biomolecular research

  • Boltz-2

    Accurate in silico screening for early-stage drug discovery

  • BoltzGen

    Design proteins and peptides to bind biomolecular targets

  • DiffDock

    Protein-ligand docking with diffusion models and confidence-ranked poses

  • DrugFlow

    Generate candidate ligands from a protein target and reference ligand using flow matching

  • Sequence Mutator

    Apply point mutations such as T315I to a protein sequence, checking each wild-type residue

  • ESMFold2

    Predict protein, DNA, RNA, and ligand complex structures with evolutionary scale modeling

  • Evo 2

    Genome modeling and design across all domains of life with the Evo 2 DNA language model

  • AutoDock Vina

    High performance molecular docking and virtual screening for drug discovery

  • Binding-Site Detector

    Find ligandable pockets and emit Vina-shaped docking boxes for fold-to-dock workflows

  • Protein Preparer

    PDBFixer cleanup, PDB2PQR protonation, and dock-ready structures without MD solvation

  • OpenDDE

    Predict protein, nucleic acid, and small-molecule complex structures with OpenDDE

  • OpenMM MD

    Prepare, solvate, and equilibrate protein structures with OpenMM molecular dynamics

  • MD Trajectory Analyzer

    Compute RMSD, Rg, RMSF, ligand H-bond occupancy, and representative frames from OpenMM trajectories

  • RFAntibody

    Antibody and TCR design with RFdiffusion, ProteinMPNN, and RF2 (stages or full pipeline)

  • FreeBindCraft

    Design miniprotein and peptide binders against a target structure using open-source scoring

  • GNINA

    CNN-accelerated protein–ligand docking and rescoring

  • Interaction Profiler

    Enumerate hydrogen bonds, hydrophobics, salt bridges, and π-stacking in docked complexes

  • MM-GBSA Rescoring

    End-point implicit-solvent MM-GBSA rescoring of docked poses or protein–ligand complexes

  • SurfDock

    Surface-informed diffusion docking with a reference-ligand pocket and screen-model rescoring

  • SURFMAP

    2D projections of protein surface features from PDB structures or SURFMAP matrix files

  • Thompson Sampling

    Active-learning virtual screening of un-enumerated combinatorial libraries via Thompson Sampling

Browse all 63 tools

Developers

Run it from your own code.

Jobs, workflows, and their results are all on the REST API, so a run can start from a notebook, a pipeline, or an AI client.

  • Scoped API keys

    Each key has read or write scope, an optional expiry, and an optional spend ceiling.

  • Safe retries

    Retrying a submit or a new run with the same Idempotency-Key returns the original instead of starting a second one.

  • The same workflows

    List templates, validate and estimate a spec, start runs, and download results over REST or with the Python SDK.

  • MCP for AI clients

    Connect Claude Code, Cursor, or any MCP client to the hosted CogniChem MCP server with your key. You pay your own model costs; we bill only compute.

API reference
# pip install cognichem-client
from cognichem_client import CogniChem

client = CogniChem.from_env()  # reads COGNICHEM_API_KEY

for template in client.workflows.templates():
    print(f"{template.id}: {template.summary}")

# One template's full spec: validate it, estimate it, then start a run
# with client.workflows.runs.create(...).
spec = client.workflows.template("smiles-embed-dock").spec

Security

Careful with your account and your money.

Security and data handling

Encrypted in transit

All traffic is HTTPS, with HSTS, a Content-Security-Policy, and related headers site-wide.

Scoped access

Sign-in through Supabase Auth; API keys carry their own scope, expiry, and spend ceiling.

Payments through Stripe

Card data stays with Stripe. Our servers receive only tokens and verified webhook events.

Enterprise terms

Data processing agreements, security questionnaires, and custom deployments on the Enterprise plan.

Start with a workflow, or just describe the problem.

The Basic plan is free: you pay only for what you run.

Describe your problemBrowse workflows