The AI for real science.
Turn a research question into literature, protocol, and computation steps in one workspace.
- Turn a research question into literature, protocol, and computation steps in one workspace.
- Build protocol drafts from papers, then audit assumptions, hazards, gaps, and yield limits.
- Call chemistry tools for molecular weights, GHS hazards, spectra, reaction checks, and quantum calculations.
- Run Python, Jupyter, RDKit, PySCF, and data analysis without leaving the research thread.
- Save protocols, audits, plots, PDFs, and notes into workspace files that survive the conversation.
- Keep evidence, assumptions, and tool outputs visible so results can be reviewed later.
Search PubMed, arXiv, and Google Scholar in plain language.
Build evidence-grounded protocol drafts with assumptions and sanity checks.
Resolve compounds, calculate properties, check hazards, and reason over reactions.
Proof-check procedures against yield limits, boiling points, hazards, and gaps.
Install packages, run quick commands, and execute inline calculations.
Save protocols, audits, PDFs, images, plots, and notes as durable files.
Run quantum chemistry calculations with PySCF and ASE.
Autonomous AI research agent.
Ask a research question. An AI agent reasons about your intent, selects the right tools, executes them, and evaluates the results. It repeats until it has a verified, cited answer.
Up to 10 reasoning cycles per question, with a scratchpad that tracks goals, findings, and next steps
The agent picks the right tool for each step: literature search, patent lookup, chemical analysis, or simulation
Every answer is grounded in real constraints: thermodynamics, kinetics, solubility, and molecular structure
New in Vicena 3.0: research that continues.
Version 3.0 turns Vicena from a single chat into a working research environment. The focus is simple: keep the record, continue the computation, and make tool results inspectable.
Workspace Files give protocols, audits, uploads, plots, and notes a durable place to live.
Jupyter work is targetable, state-aware, and organized around the notebook you are actually using.
Research and science tools return assumptions, provenance, gaps, and outputs the agent can reason over.
Workspace Files
Save protocols, audits, PDFs, plots, data files, and notes as normal files with stable paths.
Open Workspace FilesScientific simulations
Work in real notebooks with explicit targets, compact transcripts, visible UI state, and support files beside the notebook.
Open Scientific SimulationsLong-running chat
Queue follow-up prompts, jump to the latest result, and keep Files, Computer, and Tools open in side tabs.
Try the workspaceEvidence tools
Research, protocol, and chemistry tools expose assumptions, provenance, gaps, and structured evidence.
Browse tool catalogAutomated literature review & prior art research.
Search PubMed, arXiv, Google Scholar, and Google Patents in plain language. The AI agent iterates with refined queries across scientific and patent databases until it has comprehensive, cited results.
PubMed, arXiv, Google Scholar, Google Patents, and the open web searched in parallel
The agent evaluates results and reformulates queries until coverage is comprehensive
Every finding linked to its original source with DOIs, authors, publication dates, and patent numbers
Scientific simulations.
Scientists work in notebooks. They run simulations, plot results, and iterate. So we brought JupyterLab into Vicena. You and the AI agent work together in the same notebook, building on each other's work across sessions.
The agent builds notebooks cell by cell: writes code, checks output, fixes errors, and refines. You see every step live.
NumPy, SciPy, matplotlib, RDKit, PySCF, ASE, and more pre-installed. Need something else? Ask the agent to install it. It persists for next time.
Your notebooks, data, and installed packages persist across chats. Pick up where you left off.

Virtual lab simulation.
Simulate your protocol in a virtual lab before going to the bench. The AI validates each step against thermodynamic, kinetic, and chemical constraints.
Energy balances, equilibrium constants, and Gibbs free energy checked at every reaction step
Reaction rates, activation energies, and time-to-completion estimated from literature data
Solvent compatibility, miscibility, and precipitation risks flagged before you start
Extract experimental protocols.
Describe the experiment you want to reproduce. The AI agent extracts every variable from the literature and delivers a complete protocol where nothing is hallucinated.
Temperatures, concentrations, durations, reagents: each parameter found and cross-referenced independently
Every value scored by how many independent sources agree, so you know what to trust
PubChem and RDKit integration catches impossible reactions and validates molecular structures
AI for real chemistry.
Audit synthesis protocols before the bench. Predict reactions. Run quantum chemistry. Every finding cites its source, and when a tool has no data it says so — instead of faking safety.
Validates yield, atom balance, reagent-vessel compatibility, boiling points, and GHS hazards. Catches impossibilities before you run them.
20+ tools for reactions, properties, descriptors, spectra, solubility, and retrosynthesis. ReactionT5v2 neural predictor for patent-style coupling reactions.
DFT, Hartree-Fock, and MP2 calculations with PySCF. The AI picks the method and basis set, writes the code, runs it, returns the results.
