FOUNDERMETRICBENCH · OPEN AI EVALUATION INFRASTRUCTURE

Can an AI
defend the number?

Route a founder question. Inspect the selected tool. Change the business drivers. Reproduce the consequence. Retrieve the source.

PROTOCOLMCP 2025-11-25
TOOLS10 READ-ONLY
KNOWLEDGESEARCH + FETCH
EVALS24 GOLD CASES
PRIVATE DATAZERO

01 · PROMPT-TO-TOOL EVALUATOR

Make the AI defend its route.

MCP 2025-11-25 · LIVE CONTRACT
ROUTING CONFIDENCE98%
SELECTED TOOLget_refund_analysis
  1. 01
    Intent

    cash

  2. 02
    Matched signals

    cash, payment

  3. 03
    Required evidence

    gross collected, refunds, net collected

  4. 04
    Boundary

    Synthetic fixture only. Cite the formula and confidence.

Inspect the grounding source →

02 · COUNTERFACTUAL SCENARIO ENGINE

Change the drivers. See the consequence.

Deterministic synthetic analysis. Every output exposes its assumptions and can be reproduced by the MCP tool.

HEALTH SCORE85synthetic / 100
Scenario MRR
$897
+$200
Run-rate ARR
$10,764
+$2,400
Net collected
$1,822
+$274
Risk exposure
$372
-$324

Deterministic synthetic scenario, not a prediction or financial plan.

SEARCHStandard contract

Natural-language discovery with canonical citation URLs.

FETCHComplete evidence

Stable IDs retrieve full source-backed public records.

24 GOLD CASESPublic evaluation

Routing, calculation, retrieval, and boundary tests.

ZERO SECRETSSafe by construction

No private accounts, keys, writes, or hidden company data.

03 · MACHINE-READABLE EVIDENCE GRAPH

Every claim has a path back to proof.

01Synthetic records

Subscriptions, charges, refunds, and monthly revenue.

02Formula layer

MRR, ARR, movement, churn, collections, and risk.

03MCP contracts

Typed inputs, structured outputs, annotations, and sources.

04Founder decisions

Explainable scenarios with assumptions and boundaries.