§ Platform
The lifecycle of a machine-generated decision, covered end to end
Four assurance surfaces, one evidence model. Intelligence informs the decision, governance constrains it, verification proves the system that made it, and the receipt outlives all of them.
§ 01
The stack
Each surface stands alone; together they share one deterministic evidence substrate — the same canonical IDs, the same trust states, the same signatures.
Prove · flagship
Decision Receipts→
Every AI decision produces a signed, replayable evidence package: what was decided, on what evidence, under which policies, reproducible on demand. Anchored by the open Decision Receipt Specification v1.0.
Analyze
IntelGraph→
Governed graph intelligence over live feeds — 36,000+ entities, ATT&CK-mapped techniques, natural-language querying, TAXII/STIX 2.1 in and out. Analysis with provenance attached.
Govern
Agent Governance→
The runtime layer for agent fleets: default-deny tool gateways, memory with explicit trust states, role isolation, deterministic replay — including air-gapped execution.
Verify
Formal Verification→
Before agents coordinate in production, model-check the coordination itself: deadlock freedom, behavioral contracts, mutual exclusion — with machine-checkable certificates.
Adjacent surface
Omnimedia→
Evidence-native media operations — recording, calls, and field reporting where every artifact carries provenance from capture onward.
Build
API & Integrations→
REST APIs, TAXII 2.1 endpoint, SIEM integrations, CI/CD hooks. Receipts are issued where decisions happen — inside your pipeline, not after it.
§ 02
One evidence model underneath
The reason the surfaces compose: every artifact — a graph node, an agent action, a verification run, a media file — resolves to the same deterministic evidence identity and lands in the same signed chain.
- Deterministic evidence IDs. Identical content yields identical identifiers, everywhere — canonical JSON, content-addressed.
- Trust states, not vibes. Material moves through explicit states — raw, proposed, validated, approved — and policy decides what each state is allowed to influence.
- Policy as code. Deny-by-default rules evaluated at decision time, recorded in the receipt, exportable for your auditors.
- Signatures that outlive sessions. Ed25519-signed receipts, independently verifiable — no Summit account required.
SUM-EVID-v1-2b8d…44a7ADMIT
- artifact
- intel-assessment.draft.7
- origin
- intelgraph · feed:cisa-kev
- trust-state
- validated → approved
- policy
- data-residency · supply-chain · agent-governance
- chain
- 4 antecedent receipts · verified
See the whole stack run.
The Demo Hub walks every surface: issue a receipt, query the graph, watch a policy denial, replay a decision. No account needed.