33 methods in six technique families.
General x–y, spectroscopy (UV-Vis, fluorescence, FTIR, Raman, NMR), electrochemistry, materials (XRD, BET), imaging, and separation & MS.
Analysis, simulation and Bayesian optimization
Run smarter experiments.
AcquiLAB puts your measurements, physics-based simulations and Bayesian optimization in one workspace. Every result, measured or simulated, helps choose the next experiment.
Analyze33 methods, one upload
SimulateReactors to DFT
OptimizeBO with simulation in the loop
Write upLibrary, citations, export
Demo project, fictional data
- UploadInstrument files as exported, JASCO and similar included
- AnalyzeValues with units, uncertainties and the method version
- SimulateTest the idea in a model before using a sample
- OptimizeThe next experiment, chosen from everything so far
- Write upCite papers and results, export to Word or LaTeX
Analyze
From raw file to publication figure.
Upload a file, press Analyze, and the result opens in Plot Studio with the method and every setting recorded.
Change the figure, not the science.
Fits, baselines and peak areas run on the server and are recorded with the method that produced them.
When a run fails, it says why.
A missing column, a unit it can't read or too few points: the reason and the fix, on the result itself.
Ready for the journal.
Export vector PDF, EPS, TIFF or PNG, checked against journal sizes and colour-vision rules.
Simulate · new in V2
Simulate before you spend a sample.
Run physical and chemical models in the browser, fit them to your data, and send the results straight into optimization. Every run records the model version, its inputs and its quality checks.
- ReactorsBatch, CSTR, plug flow
- Transport & heatDiffusion, reaction–diffusion, membranes
- ElectrochemistryCottrell, Butler–Volmer, CV, impedance
- AdsorptionIsotherms, breakthrough columns
- CombustionFlame temperature, ignition delay
- MolecularxTB up to 100 atoms, DFT up to 40, water MD
- Your own modelRate equations or formulas you write
- Fit to dataRate constants with confidence intervals
Fit a model to your data
Estimate rate constants or adsorption capacity from measured curves, with confidence intervals.
Optimize in simulation
Let Bayesian optimization drive a model. Approve each run, or let the loop run on its own.
Guide lab experiments
Simulated runs join a lab campaign as cheaper evidence. The best result is always a measured one.
Optimize
Make every experiment inform the next.
Bayesian optimization builds a model from your results, estimates its uncertainty, and recommends where to experiment next, balancing promising conditions against unexplored ones.
Define your goal
Choose what to improve, set parameter ranges and limits, and mark combinations to avoid.
Learn from your results
The improved model fits each parameter separately, estimates noise, and switches to a log scale when the results call for it.
Check before you trust
The model check tells you whether its predictions hold up and whether more experiments are worth it.
One goal or several
Improve one outcome, or see the trade-offs between two to four and choose from the best compromises.
How the recommendations are made.
No black box. Every suggestion comes from a model you can inspect, and every run records the model version, its inputs and its seed.
- Model
- A Gaussian process learns the response and its uncertainty from your results, with one lengthscale per parameter and estimated measurement noise.
- Choosing the next run
- Expected improvement balances promising conditions against unexplored ones. Several objectives use hypervolume improvement and show the Pareto front.
- Model check
- Leave-one-out predictions and interval coverage tell you when the model can be trusted, and when to keep exploring.
- Simulation as evidence
- Simulated runs enter as a cheaper, separate source with its own lengthscale. The best result is always a measured one.
- Your limits
- Parameter ranges and outcome limits are respected, and you approve every experiment before it runs.
Bayesian optimization9 runs
Grid search (6 × 6)36 runs
Random sampling43 runs
Synthetic pH / temperature surface. Median of 30 random starts, 4 starting runs each, textbook Gaussian process with expected improvement. Not a claim about your system.
Try it: watch the model learn
A Gaussian process fitted to the runs so far. The band is the model's uncertainty, the lower curve is expected improvement, and the marked point is the next experiment it would suggest. Click the chart to run an experiment anywhere yourself.
Runs3
Best so far43.8 % at 25 °C
Next suggestion59 °C
Library and Documents · new
Your references, cited next to your results.
One reference library across all your projects, kept in sync with Zotero. Cite a paper and a measured value in the same sentence, and export the report to Word or LaTeX with the bibliography done.
One library, sorted by project
Add papers by DOI, PMID or arXiv ID, search OpenAlex and Crossref, or import BibTeX, RIS and CSL-JSON. Link each paper to the projects that use it.
Two-way Zotero sync
Connect once with your Zotero key. Items, collections, tags and notes move both ways.
Checked references
Every item shows whether its details were verified, and retracted papers are flagged before you cite them.
Cite papers and results
Type @ in a document to cite a paper or an analysis result. Choose ACS, RSC, Nature, APA 7, IEEE or Vancouver.
Assistant
Ask about your own results.
The Assistant works inside your project. It reads your campaigns, observations and analysis results before it answers, and shows you a plan before it runs anything.
Grounded in your project
Answers use your campaigns, observations and analysis results, with the values and units as measured.
Explains recommendations
Ask why a condition was suggested and get the reasoning in plain language.
Plan first
Before a simulation or fit, it lists the mechanism, every assumption with its source and the budget. Nothing runs until you approve.
AI agents · new
Use your own AI agent.
Connect Claude or another agent that supports MCP. It can analyse files, run optimizations and simulations, and save documents in your projects, within the permissions and run limits you set.
- run_analysis
- suggest_next_experiments
- run_simulation
- search_literature
- save_document
Analyse the three emission spectra in my upload folder and add them to Photocatalyst screening.
I ran Fluorescence / emission spectrum on each file. Emission maxima: 576, 578 and 581 nm.
- tool
- run_analysis · fluorescence_spectrum × 3
- project
- Photocatalyst screening
- runs
- 0 used · analyses are free
Example session, demo data
Solve real research challenges.
Process optimization
Yield, purity or selectivity across temperature, pH and loading.
Electrochemistry & energy
Fit voltammetry and impedance data, then tune conditions.
Materials & adsorption
Isotherms, breakthrough columns and formulation screens.
Manual
Every function, explained with examples.
Nine chapters on acquilab.io/docs, from your first upload to connecting an AI agent. Also as a PDF.
Ready to run better experiments?
Free plan available during beta. Start with one dataset, and add simulation and optimization when you need them.