the bundled guide, published in full
Ask the AI, and
your own notebook goes live.
This is the first-time setup guide that ships with the product, published exactly as it is. You can read it before buying. If these steps look like something you could not do, don't buy it. Finding that out is what this page is for.
The AI runs every command and every fiddly configuration screen. Your part is to log in to Cloudflare and Google, approve what it proposes, and at the end paste one secret value.
Three things to prepare
All of them are free to obtain. If you are missing one, the AI will walk you through getting it.
| Item | What it is for | Cost |
|---|---|---|
| Cloudflare account | Where the notebook lives (server, database, file storage) | Runs on the free tier |
| Google account | Logging in to the notebook. For a lab, use the PI's account | Free |
| Claude Code | The working AI that does the setup and the operations | Separate subscription |
Nothing passes through our servers. Lab records and attachments alike live only in your Cloudflare account.
Give this prompt to the AI
Open the unzipped folder in Claude Code and paste the text below as it is. From that point on, the AI runs the process.
Please perform the first-time setup of this electronic lab notebook.
Start by reading SETUP.md and AI_CONSTITUTION.md in full, and follow the order and the safety rules written there.
How I want you to work:
- Run every command yourself. Do not make me type commands.
- First inspect the current configuration and what is missing, then ask me any questions one at a time.
- When a Cloudflare or Google login is needed, open that screen, ask me to log in, and wait.
- Do the post-login configuration yourself in the browser (especially the OAuth setup in the Google Cloud Console). Do not make me configure it by hand.
- Never print passwords, verification codes, or the client secret into the chat. Give me a safe input field to type them into myself.
- Before anything that could incur charges (enabling R2, for example), explain what it does and get my confirmation.
- Do not tell me you are finished until the pre-launch check (npm run doctor:remote) passes.
- At the end, walk through it with me: I log in with Google in the browser, create one notebook page, draw a chemical structure, and save it.
While you work, tell me only what you are doing right now and what I need to do next.
In Claude Code you can simply say "set up the ELN" and the bundled eln-setup skill runs the same procedure.
Log in yourself, then give control back to the AI
The AI opens the screens it needs. Once you have logged in and cleared identity checks, leave the browser open and hand control back.
What you do by hand
- Log in to Cloudflare and press "Allow" on the authorisation screen
- Log in to Google (including two-factor authentication)
- If it asks to enable R2 (file storage), listen to the explanation and approve it
I'm logged in. From here, do the configuration yourself in the browser. In particular, do not make me set up the OAuth configuration in the Google Cloud Console by hand. Stop and call me only when a password, two-factor authentication, a secret value, or an approval that involves billing is required.
Passwords, two-factor codes, and payment details you type yourself. Never read them out to the AI or paste them into the chat.
Leave the Google setup to the AI; only paste at the end
To get a login screen, Google needs an "allow this app" configuration. That console is fiddly, so don't do it yourself — let the AI drive it.
Only two things come back to you
- Confirming the test users: tell the AI the email address you will log in with. If you plan to invite lab members, give it their addresses now and save yourself a trip later
- Pasting the client secret: the AI will provide a safe input field. Do not paste it into the chat
This is an app you built for yourself, so that warning is normal. Continue via "Advanced". You do not need to submit it for review or switch it to published (the product uses Google only to confirm who is logging in, and never touches your Gmail or Drive).
"I got redirect_uri_mismatch." "I got access_denied." Just relay the error text and the AI will find the cause and fix it. Don't start changing settings yourself.
Review the AI's pre-launch report
When the configuration is done, the AI runs a pre-launch check. Confirm together that nothing is still red.
Ask for these in the report
- Whether every test passed (there are 213 of them)
- Whether the pre-launch check (doctor) is ✓ on every item
- The URL where the notebook opens
- The email address that can log in (the owner)
- How to take a backup
Start using it with one red item and you will be fixing it later with real data inside. Clear it now.
Write one page and touch everything once
This part you do with your own hands. It exists so you can feel whether the tool sits right with you.
The order to check things in
- Open the URL and press "Sign in with Google"
- Create one notebook and add one page
- Open the structure editor, draw any molecule, and save it
- Add reagents to the reaction table and watch equivalents and yield fill themselves in
- Attach an image or a PDF and confirm you can reopen it
- Search for a word you just wrote
- Open the print report and confirm you can turn it into a PDF
- Press "Load presets" on the Reagents tab and the Equipment tab (this seeds 41 solvents and other starter data, which you can edit or delete freely afterwards)
Top-right menu → "Members". No invitation email is sent, so copy the text shown on screen and pass it to the person. Their address also has to be added as a Google test user, so tell the AI whenever you invite someone.
Ask the AI to "take a backup" and the bundled eln-backup skill runs. Research records are unrecoverable. Prove the procedure works while the notebook is still empty.
Only three things to do in operation
After launch you mostly just write. Three things recur, and you can hand all of them to the AI.
| Task | How often | How to ask |
|---|---|---|
| Backup | Monthly, or whatever interval you choose | "Take a backup of the notebook." |
| Applying a new version | When an update arrives | "Apply the new version." |
| Adding a missing feature | When you need it | "Add a column for ___ to the reaction table." |
213 unit tests ship with the product. Tell the AI to "run the tests and only deploy once they are green", and a machine stops broken code from reaching production.