Now Available

The Firewall for AI

Govern Every LLM Call in Your Organization

PII masking. Hallucination detection. Policy enforcement. Complete audit trail. Three lines of code.

Try It Live Start Free

Your App

PII masked

TruthVouch

AI Firewall

Policies enforced

LLM Provider

Hallucinations caught Audit logged

Drop-in Replacement

Three Lines of Code

Replace your LLM client import. Everything else stays the same.

# Python — drop-in replacement for OpenAI
from truthvouch import TruthVouchClient

client = TruthVouchClient(api_key="tv_...")
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "..."}]
)

// TypeScript SDK — Coming Q2 2026
// import { TruthVouchClient } from '@truthvouch/sdk';
// const client = new TruthVouchClient({ apiKey: 'tv_...' });
// const response = await client.chat.completions.create({...});

// C# — DI registration
services.AddTruthVouch(options => {
    options.ApiKey = "tv_...";
});

// Inject ITruthVouchClient anywhere
public class MyService(ITruthVouchClient client)
{
    var response = await client.Chat.Completions
        .CreateAsync("gpt-4", messages);
}

What It Does

Four layers of governance between your application and the LLM.

PII Masking

Automatically detect and mask personal data before it reaches the LLM. SSNs, emails, credit cards, phone numbers — all caught.

Policy Enforcement

Define governance-as-code policies. Block unauthorized models, enforce data boundaries, require human approval for high-risk queries.

Hallucination Detection

Ground LLM responses against your truth nuggets using AI-powered verification. Flag or block responses that contain fabricated facts — across all 9+ monitored models.

Complete Audit Trail

Every request, response, and governance decision is logged to a hash-chained audit trail — immutable, tamper-evident, and queryable. Maps to 50+ regulatory frameworks. Board-ready compliance reports on demand typically in under 5 minutes.

How It Works

Up and running in minutes, not months.

Install the SDK

3 lines of code, drop-in replacement for your existing LLM client.

AI traffic flows through the firewall

Every LLM call is governed in real-time — PII masked, policies enforced, outputs verified.

Full visibility and control

Dashboard shows every request, every policy decision, every governance action.

Install the SDK

3 lines of code, drop-in replacement for your existing LLM client.

AI traffic flows through the firewall

Every LLM call is governed in real-time — PII masked, policies enforced, outputs verified.

Full visibility and control

Dashboard shows every request, every policy decision, every governance action.

Try it Live

See TruthVouch AI Firewall in action — no sign-up required.

Try it Live

No Code Required

No code change? No problem.

The AI Firewall SDK governs your application's LLM calls. The VT Sentinel Agent governs your employees' AI tool usage — with zero code changes required.

A ~10MB Go binary deployed via Intune/GPO (Windows) or Jamf/Mosyle (macOS) sits as a local network proxy on each device. It intercepts ChatGPT, Copilot, Cursor, Claude.ai, and other AI tool traffic at the TLS layer — before it reaches the provider. Every prompt is DLP-scanned, every response is logged, every unapproved tool is blocked.

Both paths — SDK and Sentinel — feed the same audit trail, the same compliance evidence, the same governance dashboard.

~10MB

Go binary — no runtime, no dependencies

Windows + macOS

MSI (Intune/GPO) · PKG (Jamf/Mosyle)

100%

AI traffic visibility — governed and shadow

What Makes Us Different

They protect the prompt. We govern the pipeline. TruthVouch addresses all categories in the OWASP Top 10 for LLM Applications.

Feature	Lakera	Protect AI	DIY
Input safety (prompt injection)			Partial
Output governance (hallucination)			Partial
PII detection & masking	Partial	Partial	Partial
Truth grounding
Compliance evidence		Partial
MCP/Agent governance
Drop-in API/proxy integration
Shadow AI / workstation governance (no SDK required)

"They protect the prompt. We govern the pipeline."

Based on publicly available product documentation as of March 2026. Verify with each vendor for your requirements.

See full comparison →

Works With Your Existing Stack

OpenAI Anthropic Google Azure AWS Bedrock LangChain LlamaIndex CrewAI AutoGen LangGraph

Simple, Transparent Pricing

Start free. Scale as you grow.

Free

$0 /mo

1,000 gateway req/mo, PII masking, basic policy enforcement, 7-day audit log retention

Starter

$349 /mo

50,000 gateway req/mo, policy engine, hallucination detection, 90-day audit retention

Frequently Asked Questions

How long does deployment take?

The Governance Gateway SDK deploys in under 30 minutes — replace your LLM endpoint URL and add your API key (3 lines of code). The VT Sentinel Agent deploys to Windows devices via Intune/GPO MSI and macOS devices via Jamf/Mosyle PKG in under 1 hour. Both paths feed the same governance dashboard within 5 minutes of first traffic.

Which LLM providers does the AI Firewall support?

The Governance Gateway supports OpenAI (GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo), Anthropic (Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku), Google (Gemini 1.5 Pro, Gemini 1.5 Flash, Gemini 1.0 Pro), and Azure OpenAI — 9+ models across 4 providers. The VT Sentinel Agent intercepts any AI tool at the TLS layer regardless of provider, including ChatGPT web, Microsoft Copilot, Cursor, and GitHub Copilot.

What compliance frameworks does the audit trail support?

The hash-chained audit trail captures every AI call with model, prompt, response, user identity, timestamp, and governance decision. This provides evidence for EU AI Act Articles 9 (risk management), 11 (technical documentation), 12 (logging), and 17 (corrective actions), as well as ISO 42001, SOC 2 AI Annex, and NIST AI RMF. Board-ready reports can typically be generated in under 5 minutes.

Ready to govern your AI pipeline?

Start free — no credit card required. Full governance in three lines of code.

Start Free Book a Demo