MCP connector
Agents plug DCP straight into their toolchain — npx -y github:dhnpmp-tech/dcp-mcp, listed in the official MCP registry. Your GPU becomes a tool an agent can call.
Become a DCP provider
Your idle GPU is working capital.
Plug a spare NVIDIA or Apple Silicon rig into DCP and serve sovereign, in-Kingdom AI inference. Demand routes itself — agents and renters pay per token, settled in Saudi Riyal.
Illustrative scenario — actual earnings track real billed inference.
Turn spare cycles into monthly income.
DCP pays cost-plus: a transparent margin over your real running cost, never a flat tier. You stay in control of when your rig is online and what it serves.
- ∞Cost-plus payout in Saudi Riyal — you keep 75% of every billed token your rig serves.
- ∞No exclusivity, no lock-in. Pause or resume your rig any time from the desktop app.
- ∞Demand routes itself — agents and renters hit the API 24/7, not only office hours.
Your GPU serves agent-driven workloads.
DCP is agent-first. Demand does not wait for a human to click a button — agents rent GPUs and run inference automatically through the rails below. That is the paid work routed to your rig.
Agent self-serve onboarding
An agent gets its own key and trial with no human in the loop — it signs up, rents, and runs inference automatically. Demand arrives 24/7, not just office hours.
OpenAI-compatible API
Renters and agents hit api.dcp.sa/v1 with the OpenAI SDK they already use — zero rewrites. Every call that lands routes real, paid inference work to a rig like yours.
01
A signed heartbeat proves your rig is live before any traffic is routed to it.
02
The router polls a live reachability + inference probe — earned-online, never claimed-online.
03
Pause and resume on your terms; the scheduler stops sending work the moment you step away.
04
Inference runs inside an isolated runtime — the renter prompt never touches your host.
Earnings shown anywhere on this page are estimates, not a guaranteed payout.
NVIDIA or Apple Silicon GPU
An RTX 3060 Ti or better, or an Apple M2 Pro / M3 / M4 with 16 GB+ unified memory.
Windows, Linux, or macOS
The DCP daemon installs in minutes and runs detached in the background.
A stable connection
The rig joins the in-Kingdom WireGuard mesh; a residential line is fine.
A free provider account
Register, get a provider key, point your rig at DCP, and start earning.
See what your rig could return.
Pick a GPU class, set hours online per day and an expected utilisation. The estimate applies the 75% provider share to the published per-hour rate.
GPU model
Hours online / day8h
Expected utilisation50%
Gross / month1,440 SAR
DCP fee (25%)−360 SAR
1,080SAR / month — you keep
Estimate only · not a guaranteed payout
Point your rig at DCP
Run the installer on Linux, or grab the desktop app for Windows and macOS. Your real provider key is minted in the setup wizard.
curl -fsSL https://dcp.sa/install.sh | sudo bash -s -- --token <YOUR_INSTALL_TOKEN>01
Install the daemon
Download the DCP desktop app or run the one-line installer. It detects your GPU and configures everything.
02
Get verified online
Your rig joins the mesh, passes a live inference probe, and appears in the marketplace as earned-online.
03
Earn in Riyal
Renters and agents route paid inference to your rig. Only successful work is billed; you keep 75%.
Ready when you are
Earn from your GPU.
Register as a provider, point your rig at DCP, and start serving sovereign AI inference for Saudi Riyal.
Payouts in Saudi Riyal on the published revenue-share schedule.