What you get
- The Studio page and its engine. Describe, review, deploy. Triggers, conditions, loops, approvals, waits and the 33 core Flow Designer actions.
- Checks against your instance. Tables, fields and choice values are read from the instance before anything is built, so a wrong field name stops the build with the reason.
- Review before deploy. You see the generated source and a hash of everything that will be installed. Deploy installs exactly that, and stops if anything changed.
- Flows land inactive. You test and activate them in Workflow Studio yourself.
- Your choice of model. Claude, an OpenAI-compatible API, Gemini, or a model on your own machine through Ollama.
- A command line. The same build and deploy from a terminal or a pipeline.
- The source. Readable Node.js, a Dockerfile, and every later version from this page.
What you need
- Node.js 20 or newer on Windows, macOS or Linux.
- A ServiceNow instance and a login that can install applications. A developer instance with the admin account works.
- An application on that instance for the flow to go into. Pillwright Studio adds flows to an application. It does not create one.
- One AI model. An API key from Anthropic, OpenAI or Google, or Ollama with a model pulled.
- Internet for the first install.
npm installdownloads the ServiceNow SDK from npm.
A flow is hard work for a model. Small local models often fail the checks, and the build then stops and lists why. Use the largest model your hardware runs well.
Run it on your own computer
- Download the zip with the button above. Sign in, buy once, download any time. The zip is made for your account and carries a
license.jsonwith your email. - Unzip it to a permanent folder, for example
Documents\Pillwright Studio. - Open a terminal in the
pillwright-studiofolder, the one that containspackage.json. - Install.
npm installThis takes a few minutes the first time. - Create your settings file. Copy
.env.exampleto.envin the same folder. - Choose a model in
.env. For Claude, setAI_PROVIDER=anthropicandANTHROPIC_API_KEY. For a local model, setAI_PROVIDER=ollamaandOLLAMA_MODELto a model you pulled.INSTALL.mdin the zip lists all four providers. - Optional: add your instance to
.envso the page connects when the server starts. Otherwise you type it in the page. - Start it.
npm startThe terminal prints the address, the model and the instance it found. - Open
http://localhost:4600, test the connection, check the application scope, describe the flow and press Build preview. The first build is slow because it sets up an SDK project.
On your own computer the server listens on 127.0.0.1 only. Nothing else on your network can reach it.
Run it in a cloud you control
- Build the image from the folder.
docker build -t pillwright-studio . - Deploy it to Google Cloud Run, AWS App Runner, Azure Container Apps or a virtual machine. One instance, at least 2 GB of memory, a 15 minute request timeout.
- Set
STUDIO_PASSWORD. The server refuses to start on a network address without it. Your browser asks for it once. - Pass the model key and the instance login as secrets in that platform, using the same names as
.env. Never bake them into the image. - Serve it over HTTPS and add your own network controls if you can, such as a VPN or an IP allow list. Anyone with the password can deploy to your instance.
A cloud server cannot reach Ollama on your desk. Use a hosted model there, or run Ollama on the same private network.
What it costs
- Pillwright Studio download, one-time
- $100
- Later versions
- included
- Per flow, per build, per month
- $0
One purchase per account. You pay your AI provider for the model calls you make, at their prices, or nothing with a local model.
Payment is handled by Stripe. The receipt comes from Stripe to the email of this account.
What leaves your machine
- Calls to your ServiceNow instance, with the login you gave.
- Your description and the table and field names needed to write the flow, sent to the AI provider you chose. With Ollama this stays on your machine.
npm installdownloads packages from npm. The page loads fonts and icons from Google Fonts and cdnjs.
Nothing is sent to us. The download has no telemetry and no update check.
If something is off
- "no AI model is configured".
.envis missing, is not next topackage.json, or has no key in it. Fix it and restart the server. - "refused that login (HTTP 401)". The username or password is wrong, the account signs in through single sign-on only, or a developer instance is hibernating and needs waking.
- "instance URL must be https://name.service-now.com". A custom domain is not accepted. Use the instance's original service-now.com address, which still works beside the custom one.
- "now-sdk is not installed".
npm installdid not finish. Run it again and read its output. - The first build seems stuck. It is installing the SDK into the workspace folder. Allow five minutes on a slow connection.
- The build stops with a list of reasons. The model wrote something the checks refused. Name the exact table and fields in your description, or switch to a stronger model.
- Lost the zip. Sign in here and download it again. It is free after the first purchase.
Questions: info@nowidevops.com. The license terms are in LICENSE.txt inside the zip: the copy is for you or your organisation and must not be passed on.
Sales (admin)
Loading…
Open the full admin dashboard for licenses, purchases, downloads and billing problems.
| Purchased | Downloads | Last download | Last version |
|---|
Recent downloads
| When | Version |
|---|
Billing problems
| When | Type | Status | Account | Stripe session |
|---|