untangle.bio
Reactor
Centrif.
Single-pass UF
IEX column
Product tankReactorCentrif.Single-pass UFIEX columnProduct tankDefine a feed — or ferment it first. The generator lays out every viable downstream train and reports yield, purity and COGS for each route, streaming results as it finds them.
Try it free → no install · runs in the browserA diversity-preserving genetic search proposes downstream trains; a rules engine drops the ones that would never run in a plant.
Every candidate scored on recovery, purity and cost. Sort or filter on any of them.
The bioreactor is a dynamic Monod/OTR integrator, not a yield spreadsheet: growth, oxygen transfer, heat duty and titer come out of the kinetics, hour by hour.
Pick an organism from the library or set your own kinetics. The broth the fermenter produces is the real feed the downstream search starts from.
An industry-standard capital ladder with per-item installation, itemised operating cost, and a year-by-year cash flow behind every headline number — benchmarked against 61 published TEAs, replicated and costed.
One economic basis feeds every tier. NPV, IRR, payback and MSP come from the same year-by-year schedule, so NPV at the MSP is zero by construction.
300+ curated molecules with full properties (MW, pKa, solubility, pI, diffusion), or define your own. Run essentially anything.
Stream-level concentrations, flow and pH update as you build the train — and the plant boundary closes to under 1%.
Co-purify several targets with branching outlets and per-product yields and titers.
62 constraints block infeasible designs: chromatography prereqs, fouling, drying order — and models refuse rather than guess when data is missing.
Every route in yield × purity × cost space. Pick the corner you want to live in.
A printable report: AACE class band, MSP headline, capital ladder, itemised opex, year-by-year cash flow, assumptions register and warnings.
untangle ships an MCP server, so Claude or any MCP-capable assistant can call the engine the canvas runs: route generation, fermentation, simulation and TEA. You authorise it once with OAuth; inference runs on your own model and the connector only answers tool calls. Before it generates routes or prices anything, the server has your assistant confirm the inputs (molecule, scale, titer and yield, facility grade, cost basis) rather than fill in defaults. The connector comes with the Academic and Pro plans; setup is in the docs.
Ranked routes with yield, purity and CAPEX, the best one drawn as a flowsheet, and a link that opens your pick on the canvas.
The broth it makes, with concentration, oxygen and heat-duty charts over the run.
The operating settings that get there or, failing that, the titer at which the target becomes reachable.
P10, P50 and P90 COGS, a tornado of the inputs that move it, and a CSV of the balance and costs.
Flow, pH and components — or simulate the fermentation and start from the broth it actually makes.
Set yield and purity targets, then explore thousands of priced downstream sequences.
Compare by yield, purity, COGS and MSP. Drop the winner onto the canvas and print the dossier.
no install · no spreadsheets · runs in the browser
Open the app →