You describe the experiment
Say it the way you would to a colleague. No scripting, no tab-switching.
TuringDNA is the operating system for genetic engineering. Describe the experiment; it designs the variants, guides, plasmid and primers.
Supported by
The conversational engine
Say it the way you would to a colleague. No scripting, no tab-switching.
Design, build and edit in one thread. Every result stays on its construct.
Design · Read · Build · Edit
From a target to scored variants, an annotated plasmid and guides mapped onto it, without leaving the page.
Language-model scoring and in-silico directed evolution, with host, motif and liability filters applied before the bench.
In a target sequence or PDB structure · Out ranked variants with per-position scores
Zero-shot, per-nucleotide scoring across coding, regulatory and non-coding DNA.
In a reference up to 50,000 nt · Out per-nucleotide scores
Paste or import a sequence and get an annotated plasmid, with simulated digests and gels.
In GenBank, FASTA, SnapGene · Out an annotated plasmid, digest and gel
Guides scored for on-target activity and off-target risk, with every cut site mapped onto your construct.
In the construct and the edit you want · Out scored guides, cut sites mapped
The rest of the bench, in the same workspace.



Annotation & alignment. Features, ORFs, domains, MSA.
Codon optimisation. Host tables, Golden Gate sites removed.
Protein docking. Two proteins in, the contact residues out.
Vendor ordering. Guides to IDT, Synthego or Twist.
Materials & methods. A methods paragraph on every result.
Shared projects. Invite collaborators into a project.

Evo 2 scores a reference up to 50,000 nucleotides.
ESM-2 scores every position; the 650M model runs on a GPU.
Every tool is free. Compute runs on credits.
In. GenBank, FASTA, SnapGene, raw DNA or protein, plate-reader CSV.
Out. GenBank and CSV, a methods paragraph, ordering to IDT, Synthego and Twist.
Not training data. The model is public. Your sequences are not.
Audit trail. Origin, author, parameters and parent version on every construct.
The researcher network
Meet researchers, share what you are working on, and start a collaboration.
Share the work. A result, a preprint, a question or a project card.
Follow the field. See what the people you track ship.
Talk directly. Message, comment and start a collaboration.
Explore the researcher networkCompute runs on credits. Pick how many you need.
$0
For trying the whole platform.
$15 /month
For a first real project.
$50 /month
For steady, serious work.
Let’s talk
For labs and teams.
One credit is about one everyday design run; heavier runs use more. Prices in USD, sold through our reseller Paddle, with tax shown at checkout.
Before you buy
ESM-2 runs protein scoring and Evo 2 handles DNA. A larger 650M model is available too — it runs on a rented GPU and uses more of your monthly credits. Every account can reach both models; your plan sets how many credits you get, not which models you can run.
You own everything you put in and get out. The audit trail on each construct is exportable for IP and for reviewers.
Yes. The editor and the design engine are conversational and visual.
The whole workbench is free, with 10 compute credits a month. Starter is $15/month for 75 credits, and Pro is $50/month for 250 credits with no expiry on your work and top-ups anytime. See Pricing.