Compute overview
Rent a GPU by the second · or earn from your own. Real-time worker + renter state.
Active instances
| Instance | GPU | Template | Region | Uptime | Rate | Status |
|---|---|---|---|---|---|---|
| Loading instances… | ||||||
GPU pools
Templates
My workers · earn
| Worker | Tier | Status | Rate |
|---|---|---|---|
| Loading workers… | |||
Earnings total
Earning tiers
1,000 credits ≈ $8.00 · redeem as build credits or queue a Stripe payout.
Recent receipts
Kubernetes
Managed clusters on rented GPUs.
Kubernetes
One-click GKE / EKS / k3s clusters provisioned on rented GPUs. Coming soon — join the early access list to get notified.
Redeem credits
Convert earned credits to build credits or cash.
Redeem credits
Convert earned compute credits into build credits (instant) or queue a Stripe cash payout (settled monthly). Minimum 1,000 cr per redemption.
Recent redemptions
Benchmarks
Publish your GPU's tier rating.
Benchmarks
Run the trd-worker benchmark suite to publish your GPU's tier rating. Tier rating sets your default earn rate.
Usage
Detailed per-instance and per-template breakdown.
Usage
Detailed per-hour / per-instance / per-template usage breakdown. Populates after your first rental.
Alerts
Email and webhook notifications.
Alerts
Email + webhook alerts for worker offline, instance crashed, balance low, payout failed. No alert channels configured.
Billing
Compute charges roll up to your main DCS account.
Billing
Compute charges roll up to your main DCS account on app.dcsai.ai.
Settings
Profile and sign-in are managed centrally.
Settings
Profile + sign-in settings are managed from the main DCS account on app.dcsai.ai.
Virtual Machines
Long-running VMs you have provisioned on the DCS GPU fleet. Persistent disk, root SSH, fixed monthly pricing. Use when your workload outgrows on-demand instances or needs stateful services (databases, MQ, CRON).
Launch a new VM
VM launcher ships in Phase 2
A two-step wizard (pick use-case → pick GPU → configure image, storage, SSH key, cost cap → deploy). Persistent disk, snapshot scheduling, resize without rebuild.
In the meantime, browse GPU pools →Workloads
Batch jobs, scheduled runs, and long-running agent tasks. Distinct from short-lived instances (which spin up per request) and Virtual Machines (which run continuously). Queue-based execution with auto-retry + cost cap.
Active workloads
Workload queue ships in Phase 4
Submit YAML/JSON job specs, set retry policy + cost cap, watch progress in real time. Backed by the same dispatch layer as instances but with persistence + scheduling.
View running instances →Browse agents
641 pre-built agents across 14 categories. Verified, community, and experimental tiers. Each agent is signed + auditable.
Agent catalog
Agent catalog opens in v1.1
Browse, fork, or install any of the 641 catalog agents directly inside Compute. Same source as app.dcsai.ai/agent-catalog.
Read the Agents whitepaper →Publish an agent
Build, test, sign, and ship an agent to the DCS catalog. 70% publisher revenue share on paid agents.
Publisher onboarding
Publisher flow opens in v1.1
agent.yaml validator, sample test runs, signed publish to the catalog. Stripe Connect for revenue share. Until then, use the Agent Studio.
Open Agent Studio →Audit log
Tamper-evident record of every action: deploys, stops, payouts, settings changes. R+1 timestamp + R+2 signature.
Recent activity
Live audit log streams in Phase 4
Every billable + administrative action gets an R+1/+2 signed receipt. Exportable as audit bundle for regulators, accountants, lawyers.
View Receipts in the meantime →