Status
The fleet reports itself
Each environment sends a signal when it launches and while it stays afloat. The console distinguishes fresh activity, stale signals, and compute that has been torn down.
Control plane for ephemeral stacks
Your environments, afloat only when needed.
Chalupa turns your project's Compose file into an isolated remote stack. Chalupa brings together status, estimated cost, test runs, and operational signals without turning services into public endpoints.
Service ports stay on localhostSigned push signals
Fleet / now
demo-harbor
Afloat
Latest signal
12s ago
agent connected
Compute cost
$0.04/h
estimated, not billed
Container memory
last 6 minutes
3.8 GB
Ephemeral compute
Removed when the environment is torn down.
Persistent volume
Separate 20 GiB volume; it survives.
Idle window: 27 min remainingagent v1
One operational view
What is live, what retains data, and what happened.
Cost
Estimates stay with the resource
Compare hourly compute, approximate month-to-date spend, and the observed idle window. Every figure is labeled as estimated; the provider invoice remains authoritative.
Evidence
Test runs stay in context
Duration, outcome, percentiles, and artifacts appear beside the environment that produced the run, without confusing storage with observability.
Operating model
From Compose to an observable fleet.
Declarative configuration in your project, local inference, and a control plane that receives signals. Every piece has a clear boundary.
- 01
Describe the project
A chalupa.yml selects services and overrides. Compose remains the source of truth for ports, volumes, and dependencies.
- 02
Launch the environment
Pulumi creates ephemeral compute. When persistence is declared, the volume lives in a separate stack so it survives every destroy.
- 03
Receive signals
The agent pushes signed heartbeats to the control plane. Chalupa never needs to open an SSH connection to the droplet.
- 04
Test, then tear down
Connect through a tunnel, run suites, keep the evidence, and tear down compute when you finish. Persistent data remains untouched.
Two lifecycles
Tearing down compute does not delete data.
The droplet is replaceable. The persistent volume lives in a separate stack and is removed only through an explicit, destructive action. The console keeps that boundary visible for every environment.
Understand data and seedingCompute stack
Ephemeral
- Droplet and containers
- Created with task up
- Destroyed with task down
Data stack
Persistent
- Independent block storage
- Prepared once
- Survives every task down
The console observes the idle window and tells you when compute may be ready to tear down. It does not perform that action for you.
Security by topology
Observability without opening the door.
Services bind to 127.0.0.1 on the droplet and travel through an SSH tunnel. The agent only makes signed outbound requests to the control plane.
Only SSH is exposed
Databases, queues, and internal UIs never listen on the public interface.
Push-based flow
Vercel receives heartbeats; it never needs to enter the droplet or read local Pulumi state.
Verifiable signatures
Every payload is authenticated before it can change the fleet view.
One interface, explicit boundaries
Chalupa observes. Your workflow decides.
Inferred Compose
No second manual list of ports and volumes to maintain.
Commands at hand
Every environment shows how to open its tunnel and run the seed.
Read-only control plane
This first version reports status; it does not create or destroy resources from the web.
Decoupled artifacts
Chalupa keeps references; large files stay in the storage you chose.
Start in your project
Preview first. Launch second.
The quickstart separates local validation, persistent storage, compute, and tunnels so every step is deliberate.