Months ago I bought a Samsung Chromebook Pro in case I ever got locked out of a Soviet-style apartment in Copenhagen. At the time, that came with Google AI Pro for free on my personal account. Now that I’m trying to do some personal software engineering, I thought I’d try to use that account’s free credits until I felt the need to actually pay for AI usage.
Because I am a relatively heavy user of coding agents at work, I already had some parameters for personal projects. One of them is to use Pi for coding on my desktop. Pi is far leaner with extra context, saving me money. In addition, Pi opens doors to local large language model usage, a future post.
I’m on a Windows machine so I immediately have to go with wsl -d ubuntu and then I can do npm install -g --ignore-scripts @earendil-works/pi-coding-agent. So far so good, ignore the warnings.
How to use the credit is another question. After fiddling around, here are the steps, caused because my personal @gmail.com account has the credits but my projects will be hosted on my gWorkspace @ashraful.la accounts.
- Add my
@gmail.comaccount as a billing admin to my gWorkspace’s Google Cloud by going to GCP billing and then selecting my account and then going to Account management and then editing the right-hand panel. - On my personal account, go to Google Developer Program and enable the program.
- Once enabled, go to Google Developer Program Benefits.
- Find at the bottom the $10 monthly Gen AI & Cloud credits and select the billing account from step 1 and then add.
OK that’s the bean counting. Then the Google Cloud scaffolding. The non-obvious one here is you do need to turn on the Vertex AI Platform API, that’s what Pi is actually calling.
sudo snap install google-cloud-cli --classic
gcloud auth application-default login
gcloud config set project <GPROJECT_FOR_YOUR_PROJECT>
gcloud services enable aiplatform.googleapis.com
That should … be enough actually. Pi comes with google-vertex as a provider. So no need for a ~/.pi/agent/models.json.
GOOGLE_CLOUD_PROJECT=ashrafulla-coffee GOOGLE_CLOUD_LOCATION=us-central1 pi --provider google-vertex --model gemini-2.5-pro
I tried before this to use @lhl/pi-vertex but it kept crashing asking for a "baseUrl". I think I could have used it but google-vertex is native and cheaper.
So here’s a silly little guide that you can use to try to exercise $10/month on a personal @gmail.com account with your GCP projects.