Software for the bench
Describe the lab tool you need. Get a working, validated app back.
LabWren turns a scientist’s plain-language description of a lab task into a validated, ready-to-ship software tool.
- R²
- 0.999987
- Samples
- 40
- Out of range
- 2 flagged
Wells C3/D3 · 0.248 A562 → 0.237 mg/mL
solve y = -0.094626x^2 + 1.023668x + 0.010489 for y = 0.2476 -> x = 0.236813 mg/mL (in-range root of the quadratic)
| Wells | A562 | mg/mL |
|---|---|---|
| 0.063 | 0.051below LLOQ | |
| 0.248 | 0.237 | |
| 0.445 | 0.443 | |
| 0.648 | 0.664 |
The BCA starter tool on the bundled sample plate (synthetic data). Pick a sample or type a reading: every number is computed by the tool’s own analysis, not mocked up.
They get a script, not a tool.
Scientists already use ChatGPT and Claude for code, but hit a wall: they can’t run it, can’t trust it, and can’t hand it to a labmate.
Labs run on a mess of fragile spreadsheets, one-off scripts, and manual copy-paste.
What a chat assistant hands you
import pandas as pd
from scipy.optimize import curve_fit
df = pd.read_csv("plate_export.csv")
# TODO: which wells are standards?
popt, _ = curve_fit(quad, x, y)ModuleNotFoundError: No module named 'scipy'
What LabWren hands you
- Opens in a browser, with buttons and a file picker
- Shows the fit, the R², and the math for every sample
- Checked against the example file you uploaded
- One link or one file your labmate can use today
How it works
- 01
Describe
Describe the task and upload a real example file (plate reader export, instrument CSV, Excel sheet).
- 02
Build
LabWren asks the clarifying questions a good lab engineer would, then generates the tool.
- 03
Verify
The tool is checked against your example data, and the math is shown so you can trust it.
- 04
Ship
Get a tool the whole lab can use: a share link, or a single HTML file that opens in any browser. A desktop app is in development.
What we believe
Built for the bench scientist — grad student, postdoc, research associate, lab manager — who knows exactly what the analysis should do but isn’t a programmer.
The PI, core facility director, or R&D ops lead who wants reproducible analysis across the lab.
- Tools, not code.
- The output is something you click on, not something you run from a terminal.
- Trust is the product.
- A tool that’s wrong once never gets used again. Every tool shows its math, runs against real example data, and flags anything suspicious.
- Domain knowledge beats generality.
- LabWren should know what a BCA assay is, what a 96-well plate looks like, and what a SpectraMax export looks like, before the user explains it.
- Works in real labs.
- Lab computers are locked down, sometimes offline, and often run old Windows. Tools must run where the science happens.
- Data stays private.
- Unpublished research data is sensitive. Handle it that way from day one.
LabWren is in early development.
v0 is an early alpha that the founders are testing themselves. Join the waitlist and we’ll invite you when the first pilot opens.