webtunel-gpu dashboard metrics
Marketplace-sourced GPU capacity

The cheapest GPU capacity, on one platform

We continuously search the open GPU marketplace for the lowest-priced hosts that meet your hardware requirements, then run your Docker workloads on them. You get one account, one prepaid balance and one invoice - instead of a contract, a console and a bill for every provider.

Task intakePOST /v1/tasksOne API path for dashboard and SDK submissions.
Credit ledgerReserve -> settleHold estimated spend before a worker starts.
Runtime imageDocker workersBring your own image, env vars, GPU class, and timeout.
Live feedbackLogs + artifactsTask events stream back into the account dashboard.
Runtime lifecycle

From payload to paid usage without manual GPU operations.

01Validateauth + payload

The portal confirms the user, API key scope, task payload, and idempotency key.

02Reservecredit before compute

Estimated spend is locked before the runtime leases a worker or starts a GPU.

03Runworker image

Your Docker image claims the task, pulls scoped media, and streams logs.

04Settleusage ledger

Runtime seconds, artifacts, and final spend are posted back to the dashboard.

The problem

Raw GPU marketplaces are operational debt

  • You juggle SSH keys, instance lifecycles, and manual teardown to avoid idle spend.
  • Private media has to be handed to untrusted workers as raw URLs or bucket credentials.
  • There is no per-account credit control, task history, or billing surface out of the box.
The Webtunel layer

Webtunel adds the account and runtime layer

  • A queue reserves credit, leases tasks, and records every event and artifact.
  • Stream keys scope private media access to one job and expire quickly.
  • Auth, credit, invoices, and logs live in one Supabase-backed portal.
Platform architecture

One product, two surfaces, real GPUs.

The dashboard on webtunel.com drives the runtime on super-proxy.webtunel.com, which boots and controls GPU workers across multiple capacity sources.

Account & dashboardwebtunel.com

Register, add credit, submit tasks, and read timelines, logs, and invoices.

Runtime & APIsuper-proxy.webtunel.com

Queue, worker leases, stream keys, and GPU provider control plane.

GPU executionGPU workers

Your Docker worker image claims tasks and writes artifacts back.

Live control pathwebtunel.com -> super-proxy.webtunel.com -> GPU workers
Account portalAuth, credit, profile

User owns tasks, stream keys, events, invoices, and credit balance.

Runtime APIQueue, reserve, lease

The runtime validates credit, creates reservations, and assigns work.

Worker imageRun, log, artifact

Docker workers process payloads and stream task events back.

reserve.created$0.42 estimated
worker.claimedrtx4090-eu-42
usage.settled40m 40s runtime
Capabilities

Everything a GPU processing account needs.

Queue, credit, keys, logs, instances, and billing - all owner-scoped.

Task queue

Submit jobs over REST with idempotency keys. The runtime leases work to workers and tracks every state change.

Credit enforcement

Credit is reserved before a task runs and settled against real instance usage. No silent overspend.

Stream keys

Workers get short-lived, job-scoped keys to private media instead of raw storage URLs or bucket credentials.

Live logs

Stream worker stdout and runtime events into the dashboard while a task is still in flight.

Instance usage

See which GPU instances are running, GPU class, hourly rate, and utilization in one table.

Admin credits

Operators grant, adjust, and audit credit balances with a full transaction ledger.

Billing & invoices

Stripe-backed credit packages, downloadable invoices, and a running spend breakdown.

Profile & data

Owner-scoped rows keep every task, key, and invoice tied to the account that created it.

Use cases

Built for production GPU pipelines.

Video processing

Transcode, segment, and analyze private video without exposing source URLs to GPU workers.

Computer vision

Run detection and segmentation models across batches of frames with per-second billing.

Batch inference

Fan large prompt or embedding jobs out to warm workers and collect artifacts per task.

Media pipelines

Chain extract, enrich, and render stages where each stage is its own queued task.

Internal GPU jobs

Give internal teams a credit-metered, account-scoped surface instead of raw provider access.

Pricing & credit model

Prepaid credit, marketplace GPU rates.

Top up credit, reserve it before a task runs, and pay by real instance usage. Examples track live multi-source GPU market rates.

RTX 4090

24 GB VRAM

Ada
High availability
$0.34/hr

$0.28 - $0.46/hr range

Rent capacity

A100 PCIe

80 GB VRAM

Ampere
Medium availability
$1.12/hr

$0.96 - $1.48/hr range

Rent capacity

H100 SXM

80 GB VRAM

Hopper
Limited availability
$2.48/hr

$2.10 - $3.20/hr range

Rent capacity

Credit is reserved when a task is enqueued and settled against actual instance usage. Cancellation releases the unused reservation.

Developer workflow

Submit, reserve, process, inspect.

01. Submit a task

POST /v1/tasks with a payload and an Idempotency-Key from dashboard or API.

02. Runtime reserves credit

The runtime checks the balance, reserves estimated cost, and enqueues the task.

03. Worker processes

A worker claims the lease, runs your image, and reports progress and artifacts.

04. Artifacts & logs appear

Timeline events, logs, and artifacts surface on the dashboard task view in real time.

Security & operations

Least-privilege by default.

Auth, API keys, stream keys, and row ownership keep every account isolated and every worker scoped to one job.

Supabase AuthScoped keysOwner-scoped rows

Supabase Auth

Sessions, password reset, and JWT verification run through Supabase Auth.

Scoped API keys

Keys carry least-privilege scopes for task submission and runtime calls.

Short-lived stream keys

Media keys are minted per job with a TTL and revoked on cancel.

Owner-scoped rows

Row-level ownership keeps tasks, keys, and invoices isolated per account.

Resources

Go deeper before you build.

Docs

Documentation hub

Auth, credits, task flow, stream keys, and the worker image tutorial.

Open docs
Resources

Guides & references

Implementation guides, architecture references, and operational checklists.

Browse resources
Blog

Platform engineering

How credit-safe queues, the runtime split, and stream keys are built.

Read the blog
Ready for real workloads

Bring a Docker image, send a task, track every dollar.