How to run
Google Cloud locally.

Hitting real GCP during development means a project, billing, credentials and a network. Running it locally removes all four. There are two common ways to do it.

Your options at a glance

• The gcloud emulators — Google ships a separate local emulator per service (Pub/Sub, Firestore, Datastore, Bigtable, Spanner, Storage). Official and accurate, but each is its own process and port, and they cover GCP only. (Vyomi vs the GCP emulators →)
• Vyomi — the Multi-Cloud Digital Twin Runtime. Runs GCP plus AWS and Azure behind one endpoint, on real backends, with native consoles and Terraform export. (See the full landscape →)

This guide uses Vyomi, because the same steps then work for AWS and Azure too.

Step by step

1Install & start

brew install vyomi-cloud/tap/vyomi   # or: docker run -p 9000:9000 vyomi/appliance
vyomi up

Vyomi now answers Google Cloud API calls at http://localhost:9000.

2Cloud Storage (google-cloud-storage)

from google.cloud import storage
client = storage.Client(
    project="local",
    client_options={"api_endpoint": "http://localhost:9000"},
)
bucket = client.create_bucket("demo")
bucket.blob("hello.txt").upload_from_string("hi from local GCP")
print([b.name for b in client.list_blobs("demo")])

3Firestore (google-cloud-firestore)

from google.cloud import firestore
db = firestore.Client(project="local",
        client_options={"api_endpoint": "http://localhost:9000"})
db.collection("users").document("u1").set({"name": "Ada"})
print(db.collection("users").document("u1").get().to_dict())

4Drive it with Terraform

Point the Google provider at the local endpoint, then terraform apply. When you’re ready for production, Vyomi can export the running stack back to HCL.

provider "google" {
  project = "local"
  storage_custom_endpoint = "http://localhost:9000/storage/v1/"
}

Going multi-cloud

Because Vyomi runs AWS and Azure from the same appliance, the identical pattern — change the endpoint, keep the SDK — works for boto3 and azure-storage-blob too. That’s the difference between a per-service GCP emulator and a full multi-cloud digital twin.