Vicena Logo Vicena
Protocol audits, chemistry fact packs, and PySCF notebooks
Explore protocol design and audit certificates

AI for real chemistry.

A chat interface for protocol audits, compound and reaction fact packs, and small-molecule quantum chemistry.

Calculate stoichiometry, vapor pressure, pH, and yield with explicit assumptions.

  • Calculate stoichiometry, vapor pressure, pH, and yield with explicit assumptions.
  • Build compound fact packs from PubChem, RDKit, Thermo, and SMARTS in one pass.
  • Build reaction fact packs with balance, compatibility, energetics, kinetics, and gaps.
  • Run DFT and Hartree-Fock calculations without writing PySCF boilerplate.
  • Draft a protocol from evidence, then audit it as a structured certificate.
  • Never "no hazards detected": tools return coverage gaps instead of false safety.
Built on
RDKit ·PubChem ·Thermo ·PySCF ·ChemPy ·Molecular Transformer
Build operational drafts, then audit them for feasibility, safety, and gaps.
Protocol Workflows

Build operational drafts, then audit them for feasibility, safety, and gaps.

Balance, compatibility, prediction, energetics, kinetics, and boundaries.
Reaction Intelligence

Balance, compatibility, prediction, energetics, kinetics, and boundaries.

Run DFT, Hartree-Fock, and MP2 calculations with PySCF in your notebook.
Quantum Chemistry

Run DFT, Hartree-Fock, and MP2 calculations with PySCF in your notebook.

Identity, properties, hazards, solubility, spectra, and data gaps.
Chemical Intelligence

Identity, properties, hazards, solubility, spectra, and data gaps.

Work backwards from target molecules to find feasible synthetic routes.
Retrosynthesis

Work backwards from target molecules to find feasible synthetic routes.

Validated stoichiometry, solutions, pH, vapor pressure, and gas pressure.
Chemistry Calculator

Validated stoichiometry, solutions, pH, vapor pressure, and gas pressure.

How it works

Deterministic tools, in front of the LLM.

The agent has access to chemistry tools that run deterministically (RDKit, PubChem, Thermo, PySCF, ChemPy, and reaction models). When the LLM needs a molecular weight, a boiling point, a GHS code, or a reaction product, it calls a tool and uses the result directly. Numbers in the answer trace back to the tool call that produced them.

🔬

Sources on every number

Molecular weights come from RDKit. Boiling points from the Thermo library. GHS hazards from PubChem. Every value in the answer cites the library and function that produced it, so you can verify it or rerun it yourself.

⚖️

Coverage gaps, not false safety

When a tool has no data for your compound, it returns a "coverage gap" message, never "no hazards detected". The agent then falls back to its own chemistry knowledge and labels that part as reasoning, not as a verified result.

🔗

Multi-tool answers in one prompt

A protocol audit might call yield, atom-balance, compatibility, boiling-point, and safety tools. The agent picks which tools to run, in what order, and reads the outputs to produce the final answer. You describe the problem once.

Chemistry Lab

What's in the package.

Four groups of tools, all callable by the agent in a single chat.

NEW Protocol Builder + Audit icon

Protocol Builder + Audit

Draft first, then proof-check.

Protocol Builder separates source-backed steps from assumptions. Protocol Audit formalizes the procedure, generates proof obligations, and records findings, gaps, assumptions, and provenance as a certificate.

Chemistry Calculator icon

Chemistry Calculator

Bounded quantitative chemistry.

Stoichiometry, theoretical yield, solution preparation, dilution, pH, simple equilibrium, vapor pressure, and ideal-gas pressure. Missing constants or ambiguous species produce validation errors, not guesses.

Chemical + Reaction Intelligence icon

Chemical + Reaction Intelligence

Evidence packets for agent reasoning.

Chemical Intelligence resolves compound identity, properties, hazards, solubility, spectra, and similarity. Reaction Intelligence adds reaction interpretation, prediction, balance, compatibility, energetics, kinetics, and explicit data gaps.

⚛️

Quantum Chemistry

DFT, Hartree-Fock, MP2. Run in your cloud notebook.

The AI picks the method and basis set, builds a PySCF input, executes it in your persistent JupyterLab environment, and returns energies, optimized geometries, HOMO-LUMO gaps, and spectra.

What you can get done

Five practical jobs this does for working chemists.

Each scenario below is a practical chemistry task. Read what the agent does, what comes back, and what manual work it removes.

01

Before the bench, know if your protocol actually works.

You pull a synthesis from a paper or design one yourself. You're about to commit time, reagents, and glassware. Vicena reads your protocol, checks the math, flags incompatibilities, and cites what it found, in about 30 seconds.

What you see

A step-by-step audit showing theoretical max yield vs. your claim, atom balance, boiling points of solvents vs. reflux temperatures, GHS hazards for each reagent, and compatibility concerns (bleach + ammonia, HF + glass, chlorinated solvent + polycarbonate). Every number cites the library it came from. Anything it cannot verify is marked as a coverage gap, not glossed over.

What it saves

Half an hour of manual SDS review, one "wait, is this yield actually possible?" embarrassment in group meeting, and the reagents you would have wasted running a protocol that could not work.

02

From a paper in your hand to a runnable protocol.

Papers, supporting information, and patents often bury the method in dense prose. Vicena extracts variables such as temperature, concentration, duration, and reagents, cross-references them against the source claims, and flags anything inconsistent or incomplete.

What you see

A structured protocol with every parameter labeled, a confidence score for each value, and the page/paragraph it came from. Ready to paste into your lab notebook.

What it saves

The hour you would have spent reading methods sections three times because the supplementary has different numbers than the main text.

03

Reaction products, without guessing.

For patent-common reactions such as Suzuki, Buchwald, Heck, amide coupling, and Fischer esterification, Vicena can call a trained reaction model. For textbook reactions such as SN2, acid-base, and simple esterification, it can use deterministic rules. You get the predicted product, the method used, and a clear reliability boundary.

What you see

The predicted product as SMILES and structure, the method that produced it, and where the prediction is expected to be reliable vs. where it is not.

What it saves

Mental bandwidth on the reactions where the answer is already known, so you can spend your thinking on the ones that are not.

04

Drug-likeness and scaffold analysis in one prompt.

Paste a SMILES. Get LogP, MW, HBD, HBA, TPSA, rotatable bonds, Lipinski pass/fail, Veber pass/fail, and Tanimoto similarity to the reference compounds you choose.

What you see

A full descriptor table with rule-based assessment, plus similarity scores against your reference set with scaffold analysis.

What it saves

Five tabs of SwissADME plus manual fingerprint comparison, with consistent methodology and structured output.

05

DFT without the PySCF boilerplate.

You want a HOMO-LUMO gap, an optimized geometry, a dipole moment, or a vibrational spectrum on a small-to-medium molecule. You don't want to write PySCF input files or remember basis set conventions. Ask in plain language. The agent picks an appropriate method and basis set (B3LYP/6-31G* by default for DFT, sensible defaults for HF and MP2), writes the PySCF code, runs it in your notebook, and returns the numbers.

What you see

A Jupyter notebook with the calculation, the results with proper units, and comparison against reference values where available. The notebook persists: you can modify it, extend it, or cite it in your methods section.

What it saves

The forty-five minutes to an hour you would otherwise spend on input-file wrangling for a single-point calculation.

Chemistry laboratory glassware during a synthesis reaction

Example prompts below. Click one to run it.

Example prompts

What you can ask.

Five categories, three prompts each, from quick lookups to multi-step workflows. Click any to open a fresh chat with that prompt.

🧪

Lab safety

GHS hazards, compatibility checks, and protocol-level audit.

Uses PubChem GHS + curated compatibility rules
🌿

Green chemistry

Solvent selection with safety cross-checks. Atom economy and yield analysis.

Uses LogP solvent ranking + ChemPy stoichiometry
📚

Compound intelligence

Identity, structure, physical properties, hazards, solubility, spectra, descriptors, and similarity.

Uses PubChem + RDKit + Thermo + SMARTS + Morgan fingerprints
🧬

Reaction intelligence

Forward products, retrosynthesis, balance checks, compatibility, energetics, and kinetics.

Uses ReactionT5v2 + RDKit + ChemPy + Thermo + compatibility rules
⚛️

Quantum chemistry

DFT, Hartree-Fock, and MP2 in a cloud Jupyter notebook, with practical size limits.

Uses PySCF + ASE in Scientific Simulations (Plus tier)

All examples run on the free tier, except quantum chemistry which needs the cloud notebook.

Complete workflows

Describe the problem. The agent chains the tools.

Each workflow below is a real problem chemists solve every day. Click any example to run it in a fresh chat.

🔬

Audit my synthesis protocol

Catches impossible yields, unsafe pairings, and silent hazards before you go to the bench.

Chains Protocol Audit →Safety Summary →Chemical Analyzer
🧬

Plan a synthesis for a target

Retrosynthetic routes with literature backing and predicted yields.

Chains Retrosynthesis →Literature Review →Reaction Predictor
🧪

Characterize an unknown compound

Properties, predicted spectra, and functional groups in one pass.

Chains Chemical Analyzer →Spectroscopy Predictor →Functional Groups
💊

Evaluate drug-likeness

Lipinski + Veber assessment with safety and solvent compatibility.

Chains Molecular Descriptors →Molecular Similarity →Safety Summary
💧

Find the best solvent

Ranked solvent compatibility cross-referenced with GHS hazards.

Chains Solubility Predictor →Chemical Analyzer →Safety Summary
⚛️

Run quantum chemistry

DFT, Hartree-Fock, or MP2 calculations written and executed in your notebook.

Chains Notebook →PySCF →ASE

Protocol Audit

Six checks the agent runs on a protocol.

Given a protocol, the agent breaks it into steps and calls each of the checks below. They return numerical results with citations, or "coverage gap" if the underlying database doesn't have data. The LLM reads the tool output and writes the final summary; it does not invent numbers the tools didn't return.

Yield validation

Computes theoretical maximum from reactant mass and molecular weights via RDKit. A claimed yield over 100% is flagged as a conservation of mass violation.

Atom balance

Checks that every atom in the products came from somewhere in the reactants. Catches "atoms from nowhere" errors in proposed mechanisms.

Compatibility

Curated rules + SMARTS-based classification (HF + glass, alkali + protic, chlorinated + polycarbonate, bleach + ammonia) with resolver that handles compound aliases and qualifiers.

Boiling points

Exact values from the Thermo library (DIPPR correlations, 70,000 compounds), cited by source so you know what the AI knows.

GHS hazards

Live GHS classification from PubChem: signal word, hazard codes, precautionary statements. Refreshed per query, never cached stale.

Coverage gaps

When a check has no rule or data for your compound, it says so explicitly. Never "no hazards detected". Always "I cannot verify, use your judgment."

Chemist in protective gear handling a reagent in the lab

Who typically uses these tools.

Who it's for

Who these tools are for.

For professionals and students. For bench chemists and theorists. The barrier Vicena lowers is the same in both cases: the friction between reading about chemistry and actually doing it.

Bench chemists

Catch impossible yields and unsafe combinations before running a protocol.

Protocol Audit + Safety Summary

Medicinal chemists

Evaluate drug-likeness, find analogs, plan synthesis routes.

Retrosynthesis + Descriptors + Similarity

PIs & safety officers

Review student protocols at scale, with cited safety reasoning.

Protocol Audit + GHS + Compatibility

Grad students

Predict spectra, pick solvents, understand unknown molecules.

Spectroscopy + Solubility + Functional Groups

Computational chemists

Run DFT without writing PySCF boilerplate every time.

Quantum Chemistry + Jupyter

All the tools

The chemistry toolkit, in full.

Every tool the agent can use for chemistry work. Click any tool for the full reference.

Browse all tools

Lab Simulation & Validation

5 tools
🔬

Protocol Audit

Safety, yield, and feasibility check

🔥

Reaction Energetics

Virtual calorimeter and thermal safety

⏱️

Reaction Kinetics

Rate constants, half-lives, and timing

💧

Solubility Predictor

Solvent compatibility and dissolution

📊

Spectroscopy Predictor

Predicted IR, NMR, and MS fingerprints

Computational Chemistry

13 tools

Chemistry Calculator

Quantitative chemistry calculations

Chemical Intelligence

Compound identity, properties, hazards, and fingerprints

Reaction Intelligence

Reaction prediction, feasibility, compatibility, energetics, and kinetics

🧪

Chemical Analyzer

Properties and hazards lookup

🔬

Reaction Predictor

Neural network reaction prediction

🔬

Retrosynthesis

Work backwards from target molecule

🔬

Simple Reactions

Textbook reaction predictions

🔬

Functional Group Identifier

SMARTS substructure matching

🔬

Molecular Descriptors

Drug-likeness and Lipinski analysis

🔬

Molecular Similarity

Tanimoto fingerprint comparison

🔬

Safety Summary

GHS hazard information

🔬

Molecule Converters

Name, SMILES, and CAS conversion

⚛️

Quantum Chemistry

DFT, Hartree-Fock, and post-HF calculations

Industrial chemistry scale-up facility with process piping

Open-source libraries under the hood.

Libraries used

What the tools are built on.

Open-source chemistry libraries and public databases. The specific library or database behind each answer is cited in the output.

RDKit

Industry-standard cheminformatics toolkit used at Pfizer, Novartis, and Merck.

PubChem

NIH's open chemistry database: 116 million compounds with structures, properties, and safety data.

Thermo

70,000 compounds with validated thermodynamic properties (DIPPR correlations).

PySCF

Academic quantum chemistry package for Hartree-Fock, DFT, and post-HF calculations.

ChemPy

Physical chemistry library for stoichiometry, balancing, and kinetics.

Molecular Transformer

Neural network trained on 1.2 million patent reactions (IBM / EPFL).

Common questions

FAQ.

Can the AI check if my yield is physically possible?

Yes. The theoretical yield calculator uses reactant mass and molecular weights from RDKit to compute the 100% theoretical maximum. If your claimed yield exceeds that, the tool flags a conservation of mass violation. Ask: "Is 6.2 g of product from 5 g of starting material physically possible for SMILES X → Y?"

How does it check chemical compatibility?

A curated rule set plus SMARTS-based structural classification. Catches HF + glass, alkali + protic solvent, bleach + ammonia (chloramine gas), chlorinated solvent + polycarbonate, and about 20 other common incompatibilities. When no rule matches, the tool returns a coverage gap message instead of implying the pair is safe.

Which reaction prediction model does it use?

ReactionT5v2, a sequence-to-sequence neural network trained on 1.2 million patent reactions (USPTO). Published accuracy is ~93% on patent-style organic reactions, lower on organometallics and unusual chemistries. Simple textbook reactions (esterification, SN2, hydrolysis) use deterministic SMARTS rules instead.

Does it do retrosynthesis?

Yes, using the retrosynthesis mode of ReactionT5v2. Give it a target SMILES and it proposes disconnections and starting materials. Best for drug-like organic molecules; less reliable for very large molecules, polymers, or inorganic targets.

How is this different from SciFinder, Reaxys, or IBM RXN?

Those tools are curated databases (SciFinder, Reaxys) or single-purpose reaction predictors (IBM RXN). Vicena integrates multiple open-source components (RDKit, PubChem, Thermo, ChemPy, PySCF, ReactionT5v2) through an LLM that orchestrates them. It does not replace the full literature search of SciFinder or Reaxys; it covers property lookup, prediction, audit, and quantum chemistry tasks.

Is the GHS hazard data up to date?

Pulled live from PubChem per query, not cached. You get the current signal word, hazard statements (H-codes), precautionary statements (P-codes), and pictogram codes for any compound PubChem indexes.

What quantum chemistry methods are supported?

Hartree-Fock, DFT (B3LYP, PBE, and other functionals), and MP2, run via PySCF in your cloud Jupyter notebook. ASE is available for geometry handling. CPU-only, so practical for small to medium molecules (~30 heavy atoms). Larger systems or GPU-accelerated workflows are out of scope.

Does it calculate Lipinski Rule of Five?

Yes. The molecular descriptors tool computes MW, LogP (Wildman-Crippen), HBD, HBA, and TPSA via RDKit and evaluates against Lipinski's Rule of Five and Veber's rules for oral bioavailability.

Can it predict spectra?

Functional-group-based IR, ¹H/¹³C NMR, and MS fragment prediction using a SMARTS-matched spectral database. Not quantum-mechanical; use PySCF via the notebook for first-principles IR.

What does "coverage gap" mean in the output?

It means the tool looked for the compound or reaction and could not find data for it. Rather than returning an empty result (which looks like "safe" or "no hazards"), the tool explicitly says the database does not cover this case and the LLM falls back to its own chemistry knowledge, labeled as reasoning rather than as a verified result.

From the co-founder

We're trying to reduce the friction between reading chemistry and doing chemistry. Professionals and students should be able to extract ideas from papers, design experiments, audit protocols, and run computational chemistry without learning every tool one by one.

The goal is to let researchers focus on the scientific question while Vicena handles the mechanics: which tool to call, what assumptions matter, what evidence is available, and where the result stops being verified.

Think of it like Claude Code or ChatGPT Codex, tools that let you take an idea and execute it right away, regardless of programming language or framework. We want to enable chemistry in the same style: describe the objective in natural language, let the AI pick the right tools, and keep the evidence visible.

This is a small step in our journey toward AI for real science, helping students and professionals do science more enjoyably, with less friction.

Try it on your next protocol.

Free to start. Paste a protocol and ask for an audit, look up a compound, or plan a reaction.