VerraVerra
HomeProductHow It WorksDocsDashboard
Book a DemoSign in / Sign up

AI Governance for Regulated Industries

You're already using ChatGPT or Claude.
Your compliance team needs proof it's governed.

Verra sits between your internal AI tools and the model, logging every call, enforcing your policy, and generating the audit evidence your security team requires.

SOC 2 ready · HIPAA BAA available · EU AI Act controls · No raw text stored

Request Access

We onboard teams with a short call to understand your setup. Fill in your details and we'll be in touch within one business day.

Prefer a live walkthrough? Book a 15-minute demo →

Developer? Sign up and explore the proxy directly


How It Works

One URL change, and every call goes through Verra.

Works for any existing AI integration: a chatbot, a document processor, or a workflow automation. If it calls an LLM, Verra governs it.

Your Internal AI Tools

chatbot · doc processor
workflow · assistant

every call

Verra

✓ Scan for threats & PII

✓ Enforce your policy

✓ Log for audit trail

if safe

Model Providers

OpenAI · Anthropic
Azure · Bedrock · Vertex

Your compliance & security team

Full audit log of every AI call

Policy controls for what gets blocked, flagged, or logged

Evidence export for SOC 2, HIPAA, EU AI Act

Shadow AI detection across your org

Your engineering team

One URL change, one header, no SDK required

Auto-registers your app on first call

Works with existing OpenAI / Anthropic client libraries

~70ms overhead on detected attacks


Capabilities

All the governance your compliance team needs.

The problems every compliance and security team faces when AI enters the org.

Visibility

Concern

I have no idea what our AI apps are sending or receiving.

Solution

Full receipts on every call, covering what went in, what came out, and what Verra decided. Analytics and per-app drill-down give your security team the complete picture.

Compliance

Concern

I can't prove to auditors that we're governing our AI.

Solution

Every decision logged with policy version, risk level, and trace ID. One-click evidence pack export mapped to SOC 2, HIPAA, and EU AI Act controls.

Data Protection

Concern

Our AI apps might be leaking sensitive data to the model.

Solution

PII, secrets, and confidential content scanned on input and output. Caught before it reaches the model, and caught again before it reaches the user.

Threat Detection

Concern

Prompt injection or jailbreaks could compromise our apps.

Solution

Four detectors run in parallel: pattern matching, on-device ML classifiers, embedding similarity, and LLM-judge rules. Around 70ms overhead on flagged requests, because detectors run concurrently rather than in sequence.

Shadow AI

Concern

I don't know which teams are calling AI APIs without oversight.

Solution

Unregistered AI calls are flagged automatically and surfaced in the Shadow AI dashboard, giving your security team visibility into AI usage that bypasses the proxy.

Tool Governance

SCALES TO MULTI-APP

Concern

As we add more AI apps, I need to control what each one can access.

Solution

Per-app tool whitelists as your AI footprint grows. Verra filters what the model can call. HR can't touch GitHub. Finance can't call Slack.


Integration

Your engineers deploy it in an hour.

Verra is a drop-in proxy. One URL change, one header, no changes to agent code. Your security team gets full controls without slowing engineering down.

from verra import VerraClient # Reads VERRA_PROXY_URL, VERRA_API_KEY, VERRA_ORG_ID from env client = VerraClient.from_env() callbacks = [client.langchain("my-agent")] chain.invoke(input, config={"callbacks": callbacks})

Verra auto-registers agents on first call. No manual setup required.


Audit Trail

A complete audit trail, on every call.

Every agent request is logged with risk level, findings, policy version, and trace ID. One-click evidence export for SOC 2, HIPAA, and internal audits. No raw prompt text ever stored, only a hash and metadata.

CreatedAgentActionRiskFindings
2 min agofinance-report-agblockhighPII: SSN detected · sk-* pattern
5 min agocustomer-supportallowlow-
12 min agocode-review-age...warnmedPrompt injection · confidence 0.74
1 hr agohr-assistantmaskmedPII masked: email, phone
→ See the full admin dashboard

See it running on your agents.

15-minute demo. We'll map Verra to your compliance requirements and walk through the pipeline live.

Book a DemoRequest Access

SOC 2 compliant  ·  HIPAA BAA available  ·  No raw text stored