Introduction
What VDM Nexus is, what signed inference means, and why you'd use it.
VDM Nexus is signed inference for AI agents: an inference call paired with an on-chain payment and a cryptographic receipt of exactly what the model returned.
The wedge in one paragraph: when an autonomous agent makes a decision based on an LLM call, downstream systems usually have to trust that the agent's report of "the model said X" is accurate. A signed inference returns a hash-bound receipt of the prompt and the response, signed by the inference provider, alongside an on-chain payment that proves the call was real. The agent can hand that receipt to another agent, a smart contract, or a human auditor — and they can verify it without trusting the agent.
What's live
| Surface | Description |
|---|---|
@vdm-nexus/sdk | Lean prepaid flow. Ed25519-signed requests, debited from a credits balance. Two runtime deps. |
@vdm-nexus/x402 | Pay-per-call. Every request settles an x402 USDC transfer on Solana before the model runs. |
nexus.vdmnexus.com/api/v1/chat/completions | OpenAI-shape endpoint gated by x402. Self-hosted facilitator. |
nexus.vdmnexus.com/api/v1/inference | Original signed-request endpoint for the prepaid flow. |
Today: Solana devnet only. Mainnet rolls out behind a KMS migration for the facilitator's signing key.
Pick a starting point
- New to the concept? → Why VDM Nexus
- Want the system overview? → Architecture
- Want code? → Quickstart