Early access — cascade metrics are real (derived from canonical token telemetry); the operator field is a curated seed. Learn more about the data
◈ Platform

AI Evaluation Platform — SigRank

SigRank is an AI evaluation platform for operators. Content-free, continuous, cohort-relative, and governed — the four properties an operator evaluation platform needs.

What is an AI evaluation platform?

An AI evaluation platform is a software system that systematically measures, compares, and reports AI performance. The category includes model evaluation platforms (MMLU, LMSYS Chatbot Arena), output evaluation platforms (LLM-as-judge services, human review platforms), and LLM application evaluation platforms (Braintrust, Langfuse). Each platform targets a different layer of the AI evaluation stack.

SigRank is an AI evaluation platform for the operator layer. It measures the humans driving the AI — the variable that determines whether the model you deployed is actually being driven well. No other platform covers this layer, because until now there was no privacy-preserving way to measure it. Token counts make it possible; reading prompts does not.

Four properties that define the platform

  • Content-free. The platform captures four token pillars — input, output, cache-read, cache-write — and nothing else. No prompt content is ever read or stored. Token counts are the minimal sufficient statistic for operator evaluation, and they make the platform privacy-preserving by design.
  • Continuous. Operators are scored over 7-day, 30-day, 90-day, and all-time windows, not on a one-off test. The platform runs on every session, in the background. You see your trend — improvement is visible, regression is visible.
  • Cohort-relative. Operators are ranked against the live field, not against a static threshold. Your score reflects where you sit relative to other operators right now. The field shifts as the population grows, so the ranking stays meaningful.
  • Governed. Every snapshot is ed25519-signed on-device and verified server-side. The yield metric Υ = cache_read × output / input² is computed from signed data with cryptographic provenance. The platform is an open standard, not a proprietary black box.

How to integrate

Three integration paths. The CLI: install the sigrank npm or pip package, run a coding session, and submit your ed25519-signed token snapshot. The MCP server: use the Model Context Protocol server to let AI agents query operator scores, leaderboard data, and Yield metrics programmatically. The API: access the REST API for leaderboard data, operator profiles, and aggregate stats. The platform is free for individual operators; team and enterprise tiers add private cohorts, governance, and SLA-backed verification.

Explore the category

The SigRank Index — Methodology

How operator scores are computed from four token pillars, verified server-side, and ranked. The canonical methodology behind the SigRank platform.

Developer Documentation

How to integrate the SigRank platform: CLI installation, snapshot submission, API access, and the MCP server for agent-native workflows.

The SigRank MCP Server

The Model Context Protocol server that lets AI agents query operator scores, leaderboard data, and Yield metrics programmatically.

Pricing

The SigRank platform is free for individual operators. Team and enterprise tiers add private cohorts, governance, and SLA-backed verification.

FAQ

What is an AI evaluation platform?
A software system that measures, compares, and reports AI performance. SigRank is a platform for operators — the humans driving the AI. A new category, not a competitor to model or output platforms.
How is SigRank different?
It evaluates the operator, not the model or output. Content-free, continuous, cohort-relative, and governed. ed25519-signed snapshots, the Yield metric, and no prompt content ever read.
How does SigRank protect privacy?
Token counts only — never prompt content. Snapshots are ed25519-signed on-device and verified server-side. Operators appear under codenames. Token counts are the minimal sufficient statistic for operator evaluation.
How do I integrate?
Three paths: CLI (npm/pip package), MCP server (agent-native queries), or REST API (leaderboard and stats). Free for individuals; team and enterprise tiers add private cohorts and governance.