Provenance
Context-aware identity verification with cryptographic proof, temporal momentum, and risk overlays.
Last updated Mar 6, 2026
Layer: Agent (certification layer)
Scale: 0-100 with Certified / Conditional / Uncertified
Production Tier: Transaction-Grade
Purpose
Provenance determines whether an agent identity can be trusted for operational use. It consolidates identity, deployment, and control evidence into a single confidence-bearing trust signal before certification and onboarding decisions.
How It Works
Emits
Scoring Dimensions
1. Deployment Verification
Verifies that the deployed agent matches registered identity and ownership claims.
2. Capability Attestation
Assesses evidence that declared capabilities are independently validated.
3. Version Integrity
Checks release lineage, change control, and integrity continuity across versions.
4. Behavioral History
Evaluates stability and consistency of observed behavior over time.
5. Transparency Coverage
Measures explainability and auditability depth for decisions and actions.
Public note: exact formulas, weighting logic, and calibration constants are intentionally withheld.
Input Schema
| Field | Type | Required | Description |
|---|---|---|---|
entity_id | string | yes | Stable agent identifier. |
owner | string | yes | Responsible team or principal. |
deployment_attestations | object[] | yes | Deployment and control attestations. |
capability_attestations | object[] | no | Capability validation artifacts. |
version_metadata | object | yes | Release/version lineage and integrity data. |
behavior_summary | object | no | Historical behavioral evidence summary. |
transparency_artifacts | object[] | no | Logs, traces, and explainability records. |
context_profile | string | no | Evaluation context profile. |
Output Schema
| Field | Type | Description |
|---|---|---|
framework | string | provenance |
version | string | Scoring specification version. |
entity_id | string | Evaluated entity identifier. |
score | number | Provenance score from 0 to 100. |
band | string | certified, conditional, or uncertified. |
confidence | number | Confidence in score quality (0 to 1). |
context_profile | string | Profile used for evaluation. |
momentum_state | string | Direction of trust movement (improving/stable/degrading). |
evidence_gaps | string[] | Missing or weak evidence areas. |
Score Interpretation
Worked Example
Scenario: a platform must onboard three external agents for production workflows.
| Agent | Deploy Verif. | Capability Attest. | Version Integrity | Behavioral History | Transparency | Score | Band | Decision |
|---|---|---|---|---|---|---|---|---|
| Agent A | High | High | High | Medium | High | 86 | Certified | Approve |
| Agent B | High | Medium | Medium | Medium | Low | 64 | Conditional | Approve with restrictions |
| Agent C | Low | Low | Medium | Low | Low | 38 | Uncertified | Block pending remediation |
Operational outcome:
- Agent A receives full production access.
- Agent B is limited to lower-risk scopes until evidence is improved.
- Agent C is held from production onboarding.
Illustrative note: values and scores above are example outputs for documentation only.
Use Cases
Enterprise Agent Onboarding
Certify internal and vendor agents before production access. Reduce identity ambiguity and ensure minimum evidence quality across teams.
Regulated Financial and Insurance Operations
Verify agent identity and control lineage for high-impact workflows. Apply conditional certification when required controls are partially met.
Public Sector and Defense Procurement
Evaluate third-party autonomous systems with a standardized trust contract. Preserve auditable certification records for oversight and review.
Marketplace and Partner Ecosystems
Establish a common trust gate across heterogeneous agents and providers. Improve interoperability through consistent certification outcomes.