Supply the fabric

Earn with your GPU

Turn idle hardware into a revenue-generating node. The fabric probes what you run, prices it by live demand, and settles your share in USDC — with receipts for every job.

How it works

01

Install the node agent

One command runs the CapIX listener agent on your machine. It registers with a signed registration token and reports a hardware profile — CPU cores, RAM, GPU model and VRAM — that places your node in a pricing class.

curl -fsSL https://capix.network/install-node.sh | bash

02

Pass conformance probes

The fabric continuously probes every node: health (heartbeat freshness, telemetry), attestation (runtime hardening flags), and performance (job success rate, latency, fraud heuristics). Probes maintain a 0–100 health score that directly moves your price.

03

Run scoped leases

Jobs arrive as scoped leases — digest-pinned images in containers with dropped capabilities, read-only filesystems, PID limits and no-new-privileges. Your node claims work; it never sees customer data outside the lease boundary.

04

Get paid in USDC

Each node earns a dynamic hourly price: base rate by hardware class, adjusted by live fabric demand, your health score, and your trust tier. Payouts settle in USDC after a verification window.

Trust tiers

observedEvery new node starts here — sellable, probe-scored, earning the base dynamic price.start here
verifiedSustained passing probes and valid attestation promote the node to full pricing.
stakedStake behind your node for a pricing premium and higher placement priority.
completedLong-lived nodes with completed job history earn the highest trust premium.

The honest mechanics

Quarantine is real

Nodes that fail probes repeatedly — or commit attestation fraud — are quarantined automatically: they lose sellable status, are excluded from pricing, and pending payouts are held. Release is a manual operator action, never automatic.

Price follows health

The reliability multiplier only discounts: a perfect health score pays full rate, a degraded node earns less until its probes recover. Demand surges when your hardware class is scarce and discounts when it is idle.

Isolation by default

Leases run with --cap-drop ALL, --read-only, --pids-limit and no-new-privileges on digest-pinned images. Customer workloads never get host access.

Provider waitlist

already running a node? open the provider console →