OpenToken NetworkDevelopment
Contribute model capacity
Connect a device and explicitly share eligible inference capacity for validated work.
Requires an account, a released CLI, and a compatible loopback model runtime.
Read the contribution guideReciprocal inference network
Share model capacity.Earn Credit.Use any available model.
Connect models you already run. When eligible spare capacity completes validated work, you earn Credit to use other Model Builds on the Network.
You share bounded inference capacity — never model weights, API keys, or machine access.
OT / NETWORK PATH01—03
InferenceCredit
Your deviceEligible capacitySharing is always opt-in
OpenToken NetworkValidated workPolicy · routing · accounting
Your accessAvailable Model BuildsUse Credit through one API
Choose an outcome
Start with what you want to accomplish.
OpenToken supports three paths. Check the current availability and trust boundary before signing in.
OpenToken NetworkDevelopment
Use the Network API
Use Credit to call available Model Builds through the hosted API.
Requires an API key, available Model Builds, and released Credit policy.
Read the API guideOwned devicesDevelopment
Use models privately
Reach your own models across paired devices through an end-to-end encrypted path.
Requires two same-account CLI devices and an explicitly paired Private Endpoint.
Read the Private Access guideThe exchange loop
Capacity in. Model access out.
OpenToken turns eligible spare inference capacity into useful access without transferring your models or credentials.
Share
Explicitly enable Network Sharing on a connected device. Private Access remains separate.
Earn
Eligible capacity earns Credit only when validated inference work is delivered.
Use
Spend Credit on other available Model Builds through the hosted model API.
One network, two trust modes
Choose the boundary that fits the workload.
Shared Inference and Private Access are separate products. They do not silently fall back into one another.
OpenToken Network
Exchange eligible capacity with other authenticated users through Credit-metered APIs.
Private Access
Reach your own models across paired devices through an end-to-end encrypted path.
Who can use it
Other authenticated consumers
Your paired devices
Application traffic
Visible where required to serve it
E2E encrypted between CLIs
Billing
Credit-metered
Free · no Credit
Activation
Per-device opt-in
Explicit pairing and mount
Failure behavior
Returns the Network outcome
Never becomes paid inference
Clear trust boundaries
Know where plaintext exists.
OpenToken states the privacy boundary before you choose a path. Network Sharing is designed for workloads appropriate for contributor-operated infrastructure.
Shared Inference
Requests run on a selected contributor device. OpenToken and that contributor can observe application content as required to serve it. Ordinary contributor hardware does not provide cryptographically enforced confidentiality.
Sensitive workloads
Use Private Access for your own paired devices. Native request and response content stays encrypted across the relay and never falls back to the paid Network.
Built for model clients
One hosted endpoint.Familiar API families.
The hosted surface is designed for controlled native passthrough while OpenToken owns authentication, routing, policy, and accounting.
Hosted model API
https://api.opentoken.sh/v1Endpoint availability and compatibility are published per released API.
FAQ
The important details, plainly.
Does OpenToken upload my model weights?
No. OpenToken shares bounded inference capacity, not model weights, local API keys, or raw machine access.
Is Credit money?
No. Credit is a closed-loop usage balance for consuming available inference on the OpenToken Network.
Is Network Sharing automatic?
No. It is disabled by default and must be enabled explicitly for each device. Eligibility is not guaranteed.
Can contributors see requests?
The selected contributor processes plaintext content. Sensitive workloads should use Private Access on devices you control.
Does Private Access use Credit?
No. Same-account Private Access is free, end-to-end encrypted, and never falls back to paid inference.
Connect your first device
Put spare model capacity to work.
Sign in to connect OpenToken, then decide exactly which devices may share capacity.