gpuintegrate
Interactive docs

Connect agents and provider GPUs without guessing.

Copy the right config for OpenClaw, Open WebUI, Continue, cURL, JS, and Python. Providers can follow a guided app setup from sign-in to online heartbeat.

Choose Customer or ProviderBrowse integrations
Choose your path

What are you trying to do today?

Start with the role that matches this computer. Each path ends with a result you can verify.

Customer

Connect your agent to a Provider GPU

Get one Base URL, API key, and available model, then verify a real streamed request.

  • Create and immediately copy an API key
  • Confirm credits and an available model
  • Pass the /v1/models and stream=true tests
  • Paste the config into OpenClaw or another client
Success: Request Trace shows Succeeded, a routed node/GPU, tokens, latency, and cost.
Open Customer DashboardRead Agent Connect guide
Provider

Share this computer's GPU

Use the desktop app to scan hardware, bind this computer, download a model, and start sharing.

  • Sign in inside the Provider App
  • Scan hardware and auto-bind this computer
  • Download a recommended model and run checks
  • Press Start sharing and keep the app open
Success: The app shows Online and Provider Dashboard shows a fresh heartbeat plus Eligible.
Open Provider DashboardRead Provider guide
Agent ConnectBase URL, API key, model, /v1/models, stream=true, and client snippets.Provider onboardingInstall the app, scan GPU hardware, pull models, run checks, and start sharing.TroubleshootingCommon errors with short fixes, copyable commands, and step reveal.
Quickstart checks

Before pasting into an agent, verify these two calls.

Short tests prevent confusing client-side errors.
/v1/models returns 200

The agent can discover available models and their availability.

curl "https://gpuintegrate.com/v1/models" \
  -H "Authorization: Bearer gc_live_..."
stream=true returns chunks

Agents feel responsive only when the streaming path is healthy.

POST /v1/chat/completions
{
  "model": "llama3.2:3b",
  "stream": true,
  "messages": [{ "role": "user", "content": "Say hello." }]
}
Agent Connect

Copy settings into your agent in one minute.

OpenAI-compatible clients all use the same Base URL, key, and model.
01Base URL pasted
02API key saved
03Model selected
04Streaming enabled
OpenClaw

Use gpuintegrate as an OpenAI-compatible remote GPU endpoint.

  1. Open OpenClaw model/provider settings.
OPENAI_BASE_URL=https://gpuintegrate.com/v1
OPENAI_API_KEY=gc_live_...
OPENAI_MODEL=llama3.2:3b
OPENAI_STREAM=true
Step 1 of 4 revealed. Open the next step when ready.
Provider onboarding

Follow the app flow without learning node tokens.

Interactive guide for Windows and macOS providers.
Step 1 of 5 ยท 0 complete

Sign in

Open gpuintegrate Node and sign in.

What to checkThe left sidebar shows your account and provider status.
If it gets stuckIf it still says signed out, open Provider Dashboard from the app and sign in again.
Success looks likeThe app knows which provider owns this computer.
Source documentation

Operational source documents for deeper review.

These are source-of-truth files in the repository.
Agent Connect integrationsdocs/agent-connect-integrations.md

Full snippets for OpenClaw, Open WebUI, Continue, cURL, JS, and Python.

Provider App onboardingdocs/provider-app-onboarding.md

Install app, Ollama setup, node binding, checks, sharing, and local GPU verification.

Production smoke checklistdocs/production-smoke-checklist.md

Login, credits, worker online, completion, ledger charge, provider earning, payout.

Provider checklist

Send this to a new provider.

  1. Sign in inside gpuintegrate Node.
  2. Scan GPU hardware from the app.
  3. Download a recommended Ollama model.
  4. Run readiness checks.
  5. Start sharing and confirm Provider Dashboard shows Online.
Troubleshooting

Reveal the next fix instead of reading a long wall.

Common customer and provider setup issues.
Selected issue

HTTP 401

The API key is missing, incomplete, or revoked.

Next fixCopy the full gc_live_... key from Agent Connect. Prefixes are not enough.
  1. 1

    Open Customer Dashboard -> Agent Connect.

Authorization: Bearer gc_live_...