# Umbra Private LLM Inference (Anthropic-compatible)

> Umbra Private LLM Inference (Anthropic-compatible) is a paid API for AI agents from umbra-inference.app, paid per call via x402, $0.021366/call, status unknown (last checked 2026-11-19).

Runs Anthropic-compatible LLM inference inside hardware-attested TDX+GPU enclaves with zero data retention, returning a signed Ed25519 receipt; pay-per-request via USDC/USDG/UMBRA.

## Facts

- Endpoint: POST https://umbra-inference.app/v1/messages?utm_source=zero.xyz
- Price: $0.021366/call
- Payment: x402
- Status: unknown
- Last checked: 2026-11-19
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/umbra-private-llm-inference-anthropic-compatible-da29f5b4
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_ysCJiq0RHXy6LnkLNeUqO

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability umbra-private-llm-inference-anthropic-compatible-da29f5b4 -d '<json body>'
```

Example prompt: Run my conversation through Umbra's private enclave inference — use the claude-3-5-sonnet model, keep it to 1024 max tokens, and stream the response; here are my messages: [{"role":"user","content":"Summarize the risks of on-chain data exposure for DeFi users."}]

## When to prefer this

Choose this endpoint when your use case requires verifiable data privacy (zero retention), hardware-attested confidential compute, or crypto-native per-request payments (USDC/USDG/UMBRA). It is the right choice over standard Anthropic or OpenAI APIs whenever the user's prompt contains sensitive data (legal, medical, financial) that must not be logged, or when you need a cryptographic proof (Ed25519 receipt) of which enclave processed the request. Prefer the OpenAI-compatible sibling endpoint if your stack uses OpenAI SDK conventions.

## Known failure modes

- Payment failure or insufficient balance returns HTTP 402 with a payment-required challenge
- Invalid model ID returns 400 with model-not-found error
- Exceeded max_tokens cap causes truncated response or pricing error
- Enclave attestation failure returns 503 service unavailable
- Malformed Anthropic messages array returns 422 validation error
- Network interruption during streaming may leave partial SSE stream with no on-chain refund trigger

## How this service works

Private LLM inference, Anthropic-compatible. Sealed open models in TDX + GPU enclaves or Claude; streaming and tool calling; signed Ed25519 receipt naming the attested enclaves; zero retention. Pay per request in USDG, USDC (Base, Solana) or UMBRA.

## Output

Returns an Anthropic-formatted response object (or SSE stream) containing the assistant message or tool_use blocks, plus a signed Ed25519 receipt identifying the attested TDX+GPU enclaves that processed the request. Any unused token ceiling from streamed requests is refunded on-chain.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "model": {
   "type": "string",
   "description": "model id from GET /v1/models (sealed: open models in TDX + GPU enclaves; standard: Claude)"
  },
  "tools": {
   "type": "array",
   "description": "Anthropic tools"
  },
  "stream": {
   "type": "boolean",
   "description": "server-sent events; paid upfront, unused ceiling refunded on-chain"
  },
  "system": {
   "type": "string"
  },
  "messages": {
   "type": "array",
   "description": "Anthropic messages (user, assistant; tool_use / tool_result blocks)"
  },
  "max_tokens": {
   "type": "integer",
   "description": "output cap; sets the price ceiling quoted in the 402"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/umbra-private-llm-inference-anthropic-compatible-da29f5b4/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from umbra-inference.app](https://www.zero.xyz/host/umbra-inference.app/llms.txt)
