HumanMirror Oracle · Production

Structured analysis that AI agents can call.

Send text, JSON objects, arrays of records or numeric series. Oracle returns stable structured output with profiles, trends, anomalies, data-quality signals, recommendations and remaining credits.

Read the docs OpenAPI 3.1 MCP discovery

REST API

Bearer authentication, versioned JSON responses, stable error codes and automatic credit accounting.

Remote MCP

Expose the humanmirror_oracle tool to MCP-compatible agents through Streamable HTTP.

Autonomous delivery

Stripe purchase, API-key issuance, credits, rate limits, usage logs and uptime checks run without manual fulfillment.

What it is useful for

pre-analyzing datasets before an agent reasons over them;
detecting outliers and simple trends in numeric series;
checking missing values and field coverage in JSON records;
extracting dominant themes from text;
returning a stable machine-readable analysis block to another workflow step.

One API call

curl -L https://humanmirror.fr/api/oracle/analyze \
  -H "Authorization: Bearer hm_oracle_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input":[10,11,12,50],"objective":"Detect anomalies"}'

FAQ

Is Oracle already used by third-party AI platforms?

No adoption claim is made. Oracle is publicly available, documented and registered for integration by compatible applications and agents.

Does an MCP call consume credits?

Yes. A successful humanmirror_oracle tool call consumes one credit, exactly like a successful REST analysis.

Can an agent discover Oracle?

Oracle is published in the official MCP Registry and exposes public MCP, OpenAPI and well-known discovery metadata.