OpenClaw, Open WebUI, Continue, Cursor, an SDK, or your own app.
- Base URL
- https://gpuintegrate.com/v1
Routing blueprint
OpenAI-compatible inference on provider GPUs.
gpuintegrate routes model requests from your local agent to verified provider GPUs. Your tools and files stay on your computer; streamed model responses come back through one familiar API.
Authenticates the request and selects an eligible provider node.
- Available models
- Checking capacity
Only approved nodes with a fresh heartbeat and the requested model receive work.
- Live capacity
- Checking live nodes
Tokens stream back to the agent while tools and files stay on the customer computer.
- SSE
- stream=true ยท text/event-stream
Built for developers
Use the tools you already know.
OpenClawOpen WebUIContinueCursorOpenAI SDKsCustom apps
How it works
One request. Four explicit stages.
The route is simple for customers and strict about provider eligibility behind the scenes.
Connect with one Base URL
Create an API key, select an available model, and point your OpenAI-compatible client to gpuintegrate.
Route to eligible capacity
The scheduler checks provider approval, node health, model readiness, capacity, and worker version.
Run inference on a provider GPU
The Provider App sends only model inference work to the local runtime. It does not expose the provider computer.
Stream tokens and settle usage
The response streams back to the customer, then the wallet ledger and provider earning are recorded.
Try the route
See the endpoint behavior before signing in.
The public demo is simulated, anonymous, and rate limited. A real completion requires a customer account and credits.
POST
Samplehttps://gpuintegrate.com/v1/chat/completionsResponse
Run the sample to see an OpenAI-compatible response and an anonymous route snapshot.
Model discovery
Is capacity available now?
Read the public model catalog and safe availability counts.
Cost visibility
Estimate a token budget.
- Estimated credits
- 1.35
- Credits / 1M tokens
- 1.80
Trust boundaries
Remote inference without remote computer control.
gpuintegrate routes model inference only. Customer agents do not gain access to provider machines.
Customer tools stay local
Files, shell commands, browser actions, and the agent runtime remain on the customer computer.
Providers expose inference only
Customers cannot open remote desktop, SSH, files, or arbitrary processes on provider computers.
Eligibility before routing
Only verified providers and approved nodes with a fresh heartbeat and ready model receive requests.
Ledger-backed billing
Successful usage, refunds, provider earnings, reversals, and payouts remain traceable.
No remote desktop, SSH, arbitrary customer containers, private host mounts, or cross-provider tensor parallelism.
Read the Trust CenterOne marketplace, two clear paths
Use capacity or provide it.
Use provider GPUs through one familiar API.
Get a Base URL, API key, available model, streaming test, request trace, and transparent credit usage.
Open Agent ConnectTurn idle GPU time into model capacity.
Install the Provider App, scan hardware, download a recommended model, start sharing, and track earnings.
Start provider setupOpenAI-compatible from the first request
Connect your agent to a provider GPU.
Start with the Base URL, one API key, and a model that is available now.