VeriCordon

Authorization evidence for AI agents

Know what your authorization boundary actually evaluated.

VeriCordon creates reproducible evidence for agent, MCP, and tool authorization decisions — locally and in CI.

Try VeriCordon

No hosted control plane required.
Raw audit JSONL is not uploaded by default.

AUTHORIZATION EVIDENCEJSON / v1

01 / Evidence check

id
exact_action_policy_binding
scope
supplied audit events
EVIDENCE STATEpartial
EVENTS WITH BINDING0 / 4

02 / The gap stays visible

The supplied audit events lacked action and policy binding. The report recorded that gap as partial.

  • missing evidence remains explicit
  • no successful result by absence
  • scope limited to supplied evidence
report.json · versioned evidencereport.md · deterministic Markdown
DOCUMENTED MISSING-EVIDENCE EXAMPLE

Inspect the decision. Preserve the evidence. Trust the artifact, not a dashboard.

Structured JSONReadable report

01 / The evidence gap

Authorization happened.
Can you prove what was evaluated?

AI agents increasingly perform actions with real consequences. When something is allowed, denied, or unexpectedly succeeds, an opaque decision isn’t enough.

Raw audit logs, scattered traces, and vendor dashboards leave engineers reconstructing the context. You need a record of what actually happened.

01Requested action
02Acting identity
03Target resource
04Available context
05Evaluated policy
06Resulting decision
07Evidence completeness
ONE INSPECTABLE RECORDScope recorded.

02 / How it works

From authorization execution to inspectable evidence.

01

Run

Supply an AgentFence policy and representative tool calls, locally or in CI.

policy + calls → evaluation
02

Capture

Collect the available authorization evidence for the boundary and calls actually evaluated.

evaluation → available evidence
03

Inspect

Review a versioned JSON evidence report and deterministic Markdown report, ready to preserve in CI.

evidence → reviewable artifact

03 / Deliberate by design

Built for evidence,
not another dashboard.

Local-first

Keep sensitive authorization evidence in your environment.

CI-native

Generate evidence as part of development and security workflows.

Fail visibly

Missing binding evidence stays partial or not_evaluated — never silently converted into success.

Inspectable

Machine-readable JSON and human-readable reports make behavior reviewable by systems and engineers.

Distinct evidence states. Uncertainty stays visible.

OBSERVEDPARTIALNOT_EVALUATEDFAILED

04 / The artifact is the output

Evidence that can travel
with the change.

Artifact excerpt
report.json · check excerptschema 1
{
  "id": "exact_action_policy_binding",
  "status": "partial"
}

Missing binding evidence stays visible.

CI artifactVeriCordon

Evidence that survives a failed gate

Generation
local / CI
Validation fails
artifact can be retained
Raw audit JSONL
excluded by default
report.json · versioned JSONreport.md · deterministic Markdown

Real check excerpt from the documented missing-evidence run. The status applies only to the supplied evidence. View the repository guide

05 / In your workflow

Where VeriCordon fits.

Agent releases

Attach authorization evidence to CI before shipping agent changes.

MCP integrations

Inspect evidence for MCP tool calls evaluated at the configured boundary.

Security reviews

Give reviewers a deterministic artifact instead of asking them to reconstruct decisions from raw logs.

Regression testing

Preserve reports alongside CI results so changes to authorization evidence can be reviewed.

Your boundary. Your evidence.

Your authorization evidence should not require trusting another control plane.

LOCAL + CI EXECUTION

VeriCordon is designed around local and CI execution.

Raw sensitive audit material does not need to be uploaded to a VeriCordon SaaS service just to understand an authorization decision.

A deliberate product decision.
Not a temporary limitation.

Built for the work ahead

Built for teams already giving
AI real capabilities.

VeriCordon is most useful once agents move beyond chat and begin calling APIs, MCP tools, infrastructure, repositories, internal systems, or business workflows.

AI PlatformPlatform SecurityDeveloper SecurityAgent InfrastructureMCP Infrastructure

Early access

VeriCordon is currently in early access.

Try the open-source evidence bundle in AgentFence on GitHub.

Need richer evidence workflows, policy regression capabilities, or organizational features? Tell us what authorization evidence your team needs.

Help shape what comes next.

Scope, clearly stated

A few things
worth clarifying.

Authorization evidence.
Not a security guarantee.

No. VeriCordon produces evidence about the authorization boundary that was actually evaluated. The current evidence bundle builds on AgentFence’s evaluation and audit primitives; it does not replace the enforcement boundary.

Make authorization
decisions inspectable.

Generate evidence you can review, preserve, and reproduce.

Try VeriCordon