Cloud dev machines
for your AI agents.

hpy gives every engineer a powerful machine inside your company’s Google Cloud. Run Claude Code, Codex and Gemini CLI on it from the Mac app, a browser or your phone. It shuts down when idle, and your code never leaves your cloud.

Google Cloud · Arm and x86 machines · Claude Code, Codex and Gemini CLI

Why teams switch

Bigger machines. Safer code.
Nothing to buy.

Scale up, scale down

From 16 to 72 cores.
In about a minute.

Resize when a build or a fleet of agents needs it, then drop back down. Up to 288 GB of memory, more than any laptop sold.

Pay as you go

$0 up front.
Billed by the second.

Secure by design

Code never leaves
your cloud.

Overnight agents

Close the lid at six.
Wake up to pull requests.

One bill

Billed through Google Cloud.
Your credits apply.

How engineers work

Start on your laptop.
Pick up on your phone.

Your agents are still workingon maya-dev in your Google Cloudclaudegemini
hpy.ninjaFileMachinesWindowWed 9:41
hpy.ninja
Machines+ New machine
maya-devc4a-standard-32starting
maya-buildc4a-standard-72shut down
maya-agentsc4a-standard-16shut down
New machine
16 cores · 64 GB$0.72/h
32 cores · 128 GB$1.44/h
72 cores · 288 GB$3.23/h
Start machine
claude+runningmaya-dev · 32 vCPU · 128 GB
9:415G
Machineshpy.
maya-devrunning
2 agents working · 32 vCPU
maya-buildshut down
c4a-standard-72
maya-agentsshut down
c4a-standard-16
‹claudemaya-dev
Message claude↑
Dev server previews

Open any dev server
on your laptop or phone.

Every port gets a private link. The hpy app’s built-in VPN takes it straight to your machine. Share it with your team when you want a review.

k3f9a2-3000.localhost.hpy.ninja
Who can open it
  • M Mayamaya@acme.com · you Can open
  • S Samsam@acme.com · reviewer Blocked
  • ? Someone elseoutside acme.com Blocked
GCP control panel

Your whole Google Cloud,
in one panel.

hpy maps the Cloud Run services, jobs, schedules, builds and machines your team runs, works out what each one is built with, and puts logs and errors next to the agents that can fix them.

acme-prod us-central1 5Services3Jobs5Schedules4VMs23Builds today
checkout-api errors rising
Detected stack Node 22 · Fastify · Postgres 16 · Stripe · Sentry
p95212 ms Errors1.8% Deployed2 h ago
staginglive→ prodlive
Called bywebPub/Sub · orders-created
UsesCloud SQL · orders-dbMemorystore · cacheSecret · STRIPE_KEY
TypeError: Cannot read properties of undefined (reading 'id') webhooks/stripe.ts:88 · 12× in the last hour
Logs live
9:415G
acme-prod 411
billing-worker 3 builds failing since a91c2e0 Roll backFix it
web
checkout-api
billing-worker
search
admin
Issue tracking

Agent work, on the board
you already use.

Start an agent from an issue. hpy links the branch, PR and preview back to it, and moves the issue along as the work moves.

GitHub IssuesLinearJiraGitLabShortcutsoonAsanasoonNotionsoon
To do2
PAY-91 Pin numpy in billing-worker Start with claude →
LIN-490 Feature flag for new cart Start with codex →
Agent working2
LIN-482 Retry failed Stripe webhooks claude · maya-dev
LIN-488 Checkout load tests gemini · maya-dev
In review1
#212 Fix flaky login test PR #320 · preview ready
Done1
#207 Cache search suggestions Merged #311
The machine

Laptops top out.
hpy machines keep going.

The most popular engineering laptops next to the machines hpy runs in your Google Cloud. Same Arm architecture as Apple silicon, so builds and Docker images behave the same.

The best laptop stops here
64
128
192
288
16-core $0.72/h
32-core $1.44/h
48-core $2.16/h
72-core $3.23/h
Laptopshpy machines in your Google CloudTap a laptop to compare

Laptop

MacBook Pro 16″, maxed out

M5 Max · 128 GB · 8 TB · 18 cores

$7,349today
$204 a month over 3 years, whether you use it or not

hpy machine

c4a-standard-72

288 GB · 72 Arm cores, same architecture as Apple silicon

$0today
$3.23 an hour while it runs. $0 when it’s off.
Automatic shutdown

Pay for the hours it works.

Time15:00
Machinerunning
Spent today$8.62
c4a-standard-32 · idle timeout 30 min
WorkingIdle, about to shut downOvernight agents on spotOff, not billed
Left on for 24 hours$34.48
On hpy$15.51
Why hpy

Safer for the company.
Faster for engineers.

A laptop is a copy of your codebase that walks out the door every evening. With hpy, the code stays in your cloud and access is one switch.

Instant offboarding

Someone leaves. Their access leaves with them, in a second.

Remove a person in hpy or your identity provider and every device disconnects at once. The machine and its code stay in your cloud, ready to hand over.

MacBook Prohpy for Mac
iPhonehpy for iOS
Chromeapp.hpy.ninja
sam-dev running
SSam Leesam@acme.com
Nothing to lose

A lost laptop is just a laptop.

Local setup ~/code/api 2.1 GB .env.production ~/.config/gcloud ~/.ssh/id_ed25519
With hpy No code, keys or secrets on the device. Sign in on a new one and carry on.
Portable

Same session, any device.

Productive

Run eight agents at once. Your laptop stays cool and quiet.

Builds, tests and agents run on the machine, not on the device in front of you. Nobody waits on a spinning fan.

For CTOs and security teams

It all runs in your Google Cloud.

One Terraform setup creates an hpy project in your organization. After that, hpy holds no keys to your cloud.

Your engineers

BrowserChrome extensionMac appiPhoneAndroidChromebook
Your Google Cloud organization

hpy project

Dev machines, no public IPDisks with snapshotsSecret ManagerCloud StorageCloud NAT, outbound onlyCloud Audit Logs

Your existing projects

Cloud RunGKEDatabasesSecrets

Machines reach your projects with each engineer’s own Google identity, so agents get exactly the access the engineer has.

What hpy knowsMachine names, sizes, on or off state, and when to notify someone.
What hpy never seesCode, files, terminal output, secrets, agent sign-ins or traffic.

Approval gates. Prod deploys and IAM changes wait for a tap.

Snapshots. Every agent run can be rolled back in one tap.

Least privilege. A scoped identity per machine.

Budgets. Allowed sizes, idle timeouts and monthly caps.

Audit logs in your project, under your retention.

Nothing exposed. No public IPs, no open ports.

Pricing

Pay per active user.
Your first one is free.

hpy bills monthly through Google Cloud Marketplace for each person who used it that month. Machines are billed by Google at list price, on the same bill.

Free

For one engineer, forever.

$01 active user

Start free
  • Unlimited machines, any size
  • Claude Code, Codex and Gemini CLI
  • Browser, Chrome, Mac, iPhone and Android apps
  • Private dev server previews
  • Phone approvals and automatic shutdown

Team

For startups shipping fast.

$49per active user / month

Start with Team
  • Everything in Free, plus
  • Unlimited users
  • Share previews with your team
  • Snapshots before agent runs
  • Shared setup templates for your stack
  • Spend caps per user
  • GCP control panel and issue tracking
For security reviews

Business

For companies with IT and security teams.

$149per active user / month

Start with Business
  • Everything in Team, plus
  • SSO with Okta, Entra ID or any SAML provider
  • SCIM provisioning and instant offboarding
  • Org-wide policies for sizes, regions and approvals
  • Session audit trail and admin activity log
  • Public preview links with expiry
  • Priority support and 99.9% uptime SLA

Enterprise

For large engineering orgs.

from$249per active user / month

Talk to us
  • Everything in Business, plus
  • Private offers that draw down your Google Cloud commitment
  • Committed-use discounts
  • Dedicated support and onboarding
  • Custom SLA, security review and DPA
  • EU or US data residency for hpy’s control plane
Side by side

How hpy compares.

hpyTop-spec laptopGitHub CodespacesGoogle Cloud Workstations
Where code livesYour Google CloudOn the deviceGitHub’s cloudYour Google Cloud
BillingYour Google Cloud billApple or a resellerGitHubYour Google Cloud bill
Cost on top of computeFrom $49 per active user, first user freeHardware, upfront or leasedBuilt into a higher hourly rate$0.05 per vCPU-hour, plus $0.20 per hour per cluster
Largest machine72 cores, 288 GB18 cores, 128 GB32 coresMany Compute Engine types
Agents run after the lid closesBuilt for itNoUntil idle timeoutUntil idle timeout
Phone app with approvalsiPhone and AndroidNoNoNo
Questions

What CTOs ask first.

Which agents work on hpy?

Claude Code, Codex and Gemini CLI. Engineers sign in to each one on their machine with their own account, as they would on a laptop. hpy never handles those sign-ins.

How is hpy priced?

Per active user per month: $49 on Team, $149 on Business, and from $249 on Enterprise. The first user in every organization is free. A user counts as active in any month they start or connect to a machine. There’s no charge per machine, so engineers can run as many as they need.

How do we pay?

Through Google Cloud Marketplace, on your existing Google Cloud bill. Machines are billed by Google at list price, and your cloud credits apply to them. Whether credits also cover hpy depends on the terms of your credit program.

Can we commit annually for a discount?

Yes, on Enterprise. We set up a Marketplace private offer with a committed minimum at a discounted rate, and usage above it is billed monthly. Private offers usually count toward your Google Cloud spend commitment.

What happens if the internet drops?

The machine and its agents keep running. Reconnect and everything is where you left it. Working on the machine needs a connection.

Which clouds do you support?

Google Cloud today, with Arm and x86 machines in every region Google offers them.

Give your team cloud machines.

We’re onboarding a small number of engineering teams on Google Cloud.

Continue with Google

Sign in with your work account to join the wait list. We only see your name and email.