Design + characterization, one loop

The cloud lab for genetic circuits.

Design your circuit on a free, honest engine, submit it, and get real characterization data plus de-risking back. The engine gets smarter with every result, so you stop iterating blind.

putting the IO in bio
reef · cloud lab
Repressilator v3 · Round 2 ✓ completed
DESIGNS
12variants
5avg parts
CONTEXT
E. colihost
TXTLcell-free
ASSAY
replicates
6-ptinducer
Repressilator output · variant_07
3 reporters · mean ± SD, n = 3 · reference-normalized
RFUtime (min) →
variant_07 · characterized
Period42 ± 3 min
Amplitude (norm.)1.8 ± 0.2
Damping ratio0.21 ± 0.05
Stability > 12hpredicted
n = 3 replicates · fit R² 0.93
✓ designFree, in the browser
✓ characterizeCell-free, reference-normalized
✓ you getA full evidence package
✓ accessDesign UI or API

Pricing set with early partners · turnaround shared per project

How it works

A design–build–test loop that finally closes.

Free to design. Pay to characterize. Sharper every round.

01 · design

You state the objective

Describe the circuit in plain language or draw it. Coral Copilot compiles it, simulates the dynamics, and flags what will fail before you spend a dollar in the lab.

studio · new circuit
operator › repress GFP when arabinose is high
Coral › compiled a 2-repressor circuit
pBADaraCLacIGFP
⚠ RBS strength · unverifiedSimulate dynamics →
02 · characterize

We build it and run it

Submit to the Cloud Lab. We synthesize and characterize it cell-free, normalized to reference circuits on every run so your results compare across batches.

cloud lab · run #0117
Build · 12 constructssynthesized
Safety screenpass
Reference circuitsspiked
Characterizing
38%
03 · learn

You get an evidence package

Data plus de-risking suggestions come back: coverage, findings, a decision log. Every result calibrates the engine, so its next prediction is sharper.

evidence package · #0117
Fold-repression12.4 ± 1.8× · measured
Response time38 ± 6 min · measured
Leakiness1.7 ± 0.4% · measured
Findings3 flagged
Decision logattached
↑ n = 3 replicates · this run calibrated the engine

Two labs, one platform

Everything in silico. Everything in vitro.

Design where it's free and fast. Characterize where it counts.

in silico

Studio

The honest, dynamics-aware engine. Compile from intent, simulate the feedback and timing Cello can't, and order with a fail-closed safety gate. Coral Copilot turns intent into a circuit and asks before it assumes.

  • Intent → circuit, in plain language
  • Dynamics simulation, not just steady-state logic
  • Evidence-tiered predictions, never a black box
  • Ordering with biosecurity screening built in
Explore the Studio

in vitro & in vivo

Cloud Lab

The biofoundry. Submit a design, we build and characterize it, and return a clean evidence package. Reference-normalized for cross-batch comparability. Cell-free now, cell-based next.

  • Submit via the UI or the API
  • Reference-normalized on every run
  • An evidence package back, not just a number
  • Cell-free today, climbing the chassis ladder
See the Cloud Lab

Why it compounds

The only loop that keeps its data.

Because design and characterization live in one place, every characterized circuit comes home as a clean, labeled result. Reference-normalization makes runs comparable, and the engine's own predictions are the frame the data sharpens. The dataset can't be bought, only accumulated, so the honest engine gets more accurate with every team that builds on reef.

designcharacterizedatacalibratethe loopCOMPOUNDS

Honest by construction

Never a black box.

Evidence tiers, always shown

Every prediction is tagged for what backs it. We never fake confidence we don't have.

measuredpredictedunverified

Your designs stay yours

You keep your circuits, your data, and your IP. Neutral infrastructure. We never compete with the thing you're building.

Responsible by default

A fail-closed biosecurity gate screens every order before anything is synthesized. Table stakes, done properly.

Onboarding early partners

Bring us your circuit.

Tell us what you're building and we'll set up a characterization pilot. Early partners shape the platform and get the first calibrated predictions.

putting the IO in bio