OpenChainGraph Suite · CONTRACT A3 · arch #5

Fuzzy-Match Calibration

Note (dated 2026-08-04): a chain composes independent computations performed in an analytical order. No step's output is carried into the next step by the execution model; each step computes from its own declared inputs. This is true of every chain on this site, including chains whose per-step copy below reads as a data handoff.

Fuzzy-match FPR/recall scoring + threshold calibration on a synthetic labelled name-pair set (ART-93) -> audit receipt (cry-05). Tunes the matching engine without touching real names.

OpenChainGraph · 2 Steps cryptographic mandate Hash-Anchored §4 chain_depth:2 Client-Side · Zero PII
Chain Topology — Fuzzy-Match Calibration Scorer → Agent-Action Audit-Trail Aggregator
§4 Execution Hash · Chain Definition Anchor
execution_hash:computing…
Chain Stages · 2 Steps
1 ROOT · D0 node
Fuzzy-Match Calibration Scorer art-93-fuzzy-match-calibration-scorer
calibration grade + threshold recommendation (H1) feed the aggregator
MCP Call · score_fuzzy_match_calibration
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "score_fuzzy_match_calibration",
    "arguments": {}
  },
  "id": 1
}
https://mcp.ainumbers.co/mcp
↓
2 TERMINAL · D2 node
Agent-Action Audit-Trail Aggregator cry-05-agent-action-audit-trail-aggregator
Exports composite calibration artifact with execution_hash (H2), final stage
MCP Call · aggregate_execution_receipts
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "aggregate_execution_receipts",
    "arguments": {}
  },
  "id": 1
}
https://mcp.ainumbers.co/mcp
Export Artifacts
Download the §4 chain definition artifact (hash-anchored composite JSON) or the §13.11 W3C Verifiable Credentials view. Both derive from the chain definition — no new hash is minted. Available after hash computation.

Ask your agent

Copy this paragraph into Claude, OpenClaw, or any MCP-aware agent to run this whole workflow on the server and check one field that answers a real question.

Question: A calibration run sits three steps deep in an agent chain. What aggregate depth does its audit trail record?
Run the AINumbers MCP tool `run_chain` with {"chain":"sanctions-fuzzy-calibration"} and read `cry-05-agent-action-audit-trail-aggregator.aggregator_chain_depth`.
Re-run with `inputs` for `cry-05-agent-action-audit-trail-aggregator`, reusing the values that result echoed, with `artifacts` changed so one receipt whose chain depth is 3. Compare the same field.
Verify: call `verify_execution_hash` (https://mcp.ainumbers.co/mcp) with `claimed_hash` set to `composite_execution_hash` and the full `composite_artifact`.
Ledger, for a human re-check: https://ledger.ainumbers.co/
PII rule: send synthetic or anonymised inputs only. The MCP server runs these kernels and logs no payloads.
Chain page: https://ainumbers.co/chaingraph/chains/sanctions-fuzzy-calibration.html