ONE TOOL · ROUTE FIRST · VERIFY WHEN IT MATTERS

One tool for the whole HumanMirror boundary.

humanmirror_do is the minimal MCP entry point. Give it an intent and it routes free to the canonical HumanMirror service. Give it a supported deterministic Outcome and it can execute, independently verify the result and keep the charge only when success is proven.

Connect in 30 seconds Run live demo — free Inspect MCP endpoint Inspect manifest

1 TOOL

Unified entry

Agents load one schema instead of a large catalog. HumanMirror resolves the correct internal boundary from the declared intent.

humanmirror_do →

0 CREDITS

Routing

Intent routing is free. The response returns the canonical endpoint and the downstream payment policy before execution.

Routing manifest →

6 CREDITS

Verified Outcome

Direct deterministic Outcome execution is charged only on independently verified success. Failed execution or verification is refunded.

Outcome contracts →

Canonical routes behind one tool.

Sanitize ShieldUntrusted payloads and prompt-injection risk → /api/v1/sanitize/shield/

Purchase FirewallIntent/payment binding before an agent purchase → /api/value/bind/

Execution ReceiptPost-execution evidence → /api/x402/execution-receipt/

Zero-Cost SinkBounded context cleanup with free grace before x402 → /api/v1/sink/

Execution MarketBest admissible provider quote, non-custodial → /api/market/request/

Payment DoctorPayment-flow diagnostics and recovery → /api/payment-doctor/

Machine ResolveUnknown intent fallback → /api/machine/resolve/

Commercial logic stays explicit.

HumanMirror One never hides the target rail. Routing costs 0 Nexus credits. The Zero-Cost Sink preserves its anonymous grace path, x402 services disclose their Base/USDC price, and Execution Market keeps its non-custodial model: 0.010 USDC to publish demand, then a disclosed fee equal to 30% of verified quote savings only when a best-execution commitment is made.

The entire MCP configuration.

Remote MCP:
https://humanmirror.fr/api/one/mcp/

Tool:
humanmirror_do

Route example:
{"intent":"best execution for this machine task","dispatch_only":true}

Verified Outcome example:
{"outcome_type":"safe_preflight","input":{"command":"..."}}

Truth boundary: HumanMirror One performs canonical routing; the downstream service remains responsible for its own authorization, execution and settlement. Direct execution inside One remains limited to supported deterministic Outcome types.