graphsub.com / wiki / operational-proof-graph
Operational Proof Graph
An operational proof graph records inputs, decisions, dependencies, approvals, and outputs so teams can replay and explain complex workflows.
Definition
An operational proof graph is a system of record for how an output was produced. It captures the path from source data through tools, decisions, approvals, and final result.
Enterprise need
Teams adopting agents need to answer what changed, who approved it, which tools ran, and why an output was accepted. Logs alone rarely provide that causal map.
GraphSub capsules
GraphSub packages proof graphs into capsules for trading control, agent reliability, regulated evidence, incident dependencies, industrial traceability, and market intelligence.
Code example
Dependency blast-radius querycypher
MATCH (s:Source {id: 'SOFR:2026-07-03'})
-[:FEEDS]->(m:Model)
-[:PRODUCES]->(o:Output)
WHERE o.status <> 'replayed'
RETURN
o.id AS output,
m.version AS model,
shortestPath((s)-[*..4]->(o)) AS proof_path
ORDER BY o.priority DESCFind downstream outputs that need replay after a source changes.
FAQ
- What does a proof graph contain?
- Sources, task state, tool calls, memory deltas, approvals, dependencies, replay metadata, and outputs.
- Is this only for compliance?
- No. The same graph improves debugging, reliability, incident response, and product analytics.