Compute & Hosting API

The Unified Compute & Hosting API enables developers and product managers to manage cloud compute, AI sandboxes, and app hosting across multiple providers through a single, standardized interface. List GPU types and prices, launch and control instances, and deploy services on RunPod, Vast.ai, SF Compute, Crusoe, Nebius, AWS EC2, DigitalOcean, Linode, Scaleway, Hetzner Cloud, E2B, Daytona, Vercel Sandbox, Runloop, Render, Vercel, and more — all with one API.

What is the Unified Compute & Hosting API?

GPU clouds, AI agent sandboxes, and app-hosting platforms each have their own API, authentication, units, and lifecycle model. One provider calls a machine a "pod", another a "droplet", "server", "devbox", or "sandbox"; some quote prices per GPU-hour in cents, others per instance-hour, and some only bill per second. The Unified Compute & Hosting API normalizes these differences, so you can build once and let your customers bring the compute provider they already use.

Compute integrations represent machines as instances, the hardware and plan catalog as GPU types, and deployable apps as services with their deployments, domains, and environment variables. Lifecycle actions such as start, stop, pause, resume, restart, and terminate are updates to an instance's status.

Key Benefits for Developers

• Single Integration: Write code once to provision and manage compute across GPU clouds, sandboxes, and hosting platforms
• Normalized Data Models: Memory and disk are always MiB, GPUs map to one canonical GPU model list, and every vendor state maps to one instance status
• One Lifecycle Model: Start, stop, pause, resume, restart, and terminate through a single status update, with allowed_actions telling you what the provider supports right now
• Real-Time Data Access: Every request hits the source API live with no caching, so instance status and availability are always current
• Zero Maintenance: No need to track API versions, deprecations, or auth changes across providers

Benefits for Product Managers

• Multi-Provider Support: Let customers run workloads on the cloud they already pay for, without custom development per provider
• Price & Availability Comparison: Show GPU prices and stock across providers and regions from one catalog
• Competitive Advantage: Launch with support for major GPU clouds, sandboxes, and hosting platforms while competitors integrate one at a time
• Reduced Time-to-Market: Ship compute and deployment features in days instead of months of integration work
• Scalable Product Strategy: Add new providers as customers ask for them without significant engineering resources

Compute & Hosting Data Objects

The Compute & Hosting API provides access to the following standardized objects:

• Instances: Pods, VMs, bare-metal servers, containers, and sandboxes, with type, normalized status, allowed_actions, region, template, GPU type, GPU model and count, vCPUs, memory, disk, list price, exposed endpoints, SSH details, timeouts, and expiry. Hosting machines link to their service through service_id.
• GPU Types: The GPU and instance-size catalog, with canonical GPU model and manufacturer, the vendor SKU, GPU count and memory, vCPUs, memory, disk, architecture, prices by market and region, and availability by region. Also used as the plan catalog for hosting tiers.
• Regions: Regions and data centers, with location, country, and the GPU models offered there.
• Templates: Images, templates, and blueprints that instances start from, with default sizing, ports, start command, and build status.
• Volumes: Network volumes, filesystems, and disks, with type, status, size, region, availability zone, and the instances they are mounted on.
• Snapshots: Memory or disk snapshots taken from an instance or a volume, which new instances can be restored or forked from.
• Usage: Usage and cost records per period, with category (GPU, CPU, memory, storage, network), quantity, unit, and cost.
• Keys: SSH keys, container-registry credentials, and secrets. Secret values are write-only and never returned.
• Services: Hosted apps and projects, with type, status, runtime, region, plan, instance counts and autoscaling limits, source repository and branch, build and start commands, and public URL. Suspend or resume a service by updating its status.
• Deployments: Builds and releases of a service, with status, trigger, environment, commit, branch, URL, and timing. Creating a deployment triggers a deploy; setting its status to canceled cancels it.
• Domains: Custom domains routed to a service, with verification status and primary/redirect settings.
• Environment Variables: A service's environment variables and the environments they apply to. Values are write-only and never returned.

Common Use Cases

GPU Marketplaces & Price Comparison

Aggregate GPU types, prices, and availability across RunPod, Vast.ai, SF Compute, Crusoe, Nebius, and other GPU clouds. Help customers find the cheapest H100 or the region with capacity right now.

AI Agent Sandboxes

Give AI agents isolated environments to run code. Create, pause, resume, and terminate sandboxes on E2B, Daytona, Vercel Sandbox, or Runloop through the same instance object, with timeouts and expiry handled consistently.

Training & Inference Orchestration

Launch GPU instances for training or batch inference on the customer's own cloud account, attach volumes for datasets and checkpoints, and terminate machines when the job finishes.

Deployment & Hosting Dashboards

Build dashboards and developer tools that list services, deployments, domains, and environment variables across Render, Vercel, and other hosting platforms. Trigger, monitor, and cancel deploys from one place.

Cloud Cost Management

Pull usage and cost records alongside running instances to track spend by instance, category, and provider, and flag idle or forgotten machines.

Infrastructure Inventory & Governance

Keep an up-to-date inventory of instances, volumes, snapshots, and SSH keys across every connected provider for security reviews, audits, and cleanup.

Why Use a Unified Compute & Hosting API?

Traditional Approach:

• Build separate integrations for each GPU cloud, sandbox, and hosting provider
• Learn multiple authentication schemes, including signed requests and token exchanges
• Convert each provider's units, prices, GPU names, and lifecycle states into your own model
• Monitor and update integrations when providers change or retire API versions
• Handle provider-specific rate limits, pagination, and asynchronous operations

With Unified Compute & Hosting API:

• Integrate once with a single API that works across compute and hosting providers
• Use one authentication flow for all providers
• Work with normalized instance, GPU type, service, and deployment objects
• Control the full lifecycle through one status field
• Automatic handling of API changes, rate limits, and provider-specific differences

Real-Time, Live Data

Unlike other integration platforms that cache data or run periodic sync jobs, the Unified Compute & Hosting API provides real-time access to compute data. Every API request hits the source provider directly, so you always see the current instance status, prices, and availability.

This real-time architecture is ideal for:

• Dashboards that show live instance and deployment status
• Schedulers that need current GPU availability before launching
• AI agents that create and tear down sandboxes on demand
• Cost tools that flag running machines as soon as they start

Privacy & Security

Unified.to never stores your customers' compute data. All requests are stateless and pass through to the provider directly. Write-only values such as environment variables, registry passwords, and secrets are sent to the provider and never returned. Traffic is regionalized (US/EU/AU) to comply with data residency requirements.