Your AI agents are running.
Nobody knows what they're doing.
See every token your agents spend, block the ones that go rogue, and hand your auditor a tamper-proof log — without changing a single line of code.
Works with your stack —
LIVE DASHBOARD
See everything your agents do. In real time.
Real dashboard from Govrix Scout. Every agent call tracked, every dollar attributed, every PII instance flagged.
The Problem
Your CFO just asked why the OpenAI bill tripled this month.
You can't answer. Nobody can. Your agents are spending money you can't see, leaking data you can't detect, and making decisions you can't audit. Here's what your production logs actually look like right now.
$47K bill from one agent
A single runaway agent loop burned $47K in one night. No alert. No cap. Nobody noticed until the invoice.
PII leaked to LLM providers
Customer SSNs, emails, phone numbers flowing to OpenAI in plain text. No detection. No masking. No audit trail.
Shadow agents everywhere
Engineering shipped 12 agents. You know about 4. The other 8 are calling GPT-4 with production data right now.
EU AI Act — August 2026
€35M in penalties for non-compliant AI systems. Your agents need tamper-proof audit logs. Today they have none.
Sound familiar? This is what Govrix fixes. ↓
THE FIX
One env var. Instant visibility.
Set one environment variable. Scout intercepts every agent call — tracking costs, catching failures, flagging PII. No code changes. No SDKs. No friction.
- Per-agent cost tracking from first request
- Works with OpenAI, Anthropic, Bedrock, and more
- Auto-discovery — agents appear in dashboard instantly
Added latency: <5ms. Your agent doesn't notice.
Open source · MIT licensed · Self-hosted in 2 minutes
HOW IT WORKS
From zero to governed in minutes
No SDK to install. No agent code to change. Just point and govern.
Deploy
Docker image or single binary. No SDK, no code changes, no dependencies.
Point
Set one env var. All agent traffic routes through Govrix — zero code changes.
See Everything
Agents appear instantly. Track costs, detect PII, enforce budgets, generate audit logs.
Point
CAPABILITIES
Everything you need. Nothing you build.
v1 is live and open source — full observability, cost tracking, PII detection, and audit trails. v2 brings managed cloud, multi-provider routing, identity, and compliance reports.
Cost Attribution
Know where every AI dollar goes
Per-agent, per-model, per-day cost breakdown. Catch $47K runaway loops before your CFO does.
Agent Discovery
See every agent. Automatically.
No registration, no SDK. Agents appear in the dashboard the moment they route through the proxy.
PII Detection
Flag sensitive data in real time
SSN, credit cards, emails, phone numbers, addresses — detected inline with <0.5ms overhead. 25+ pattern tests.
Tamper-Proof Audit
SHA-256 Merkle lineage chain
Every event gets a cryptographic hash. Immutable, compliance-ready evidence your auditor will actually accept.
Protocol Native Streaming
True SSE pass-through
Works with OpenAI, Anthropic, and any compatible endpoint. No buffering. <5ms p99 added latency.
Fail-Open Design
Governance never blocks production
Proxy is stateless and fail-open. If Govrix goes down, agents keep working. Kill switch for individual agents when needed.
Multi-Provider Routing
Route across any provider
Bedrock, Azure OpenAI, VertexAI, Cohere — load balance, failover, and route by policy. One endpoint for all models.
SSO / OIDC + RBAC
Identity without the SSO tax
OIDC authentication, role-based access control, team-scoped permissions. No $500/seat SSO upgrade fee.
Webhook Connectors
Push events everywhere
Datadog, Splunk, PagerDuty, custom webhooks. Real-time alerts when budgets break or PII is detected.
Compliance Reports
Audit-ready out of the box
One-click SOC 2, EU AI Act, HIPAA, and FINRA compliance reports. Evidence packages your auditor will accept.
Real-Time PII Masking
Redact before it leaves
Mask sensitive data in-flight before it reaches upstream APIs. Emails, SSNs, credit cards redacted at the proxy layer.
A2A Identity & mTLS
Agent certificates and permissions
Issue certificates per agent, scope permissions via mTLS, and enforce agent-to-agent trust boundaries.
v1 is free forever. v2 launches when your auditor asks.
ROADMAP
v1 is live. v2 is coming.
Scout v1
Open source. Self-hosted. Production-ready. Apache 2.0 licensed.
- ✓Transparent proxy (no code changes)
- ✓Agent auto-discovery
- ✓Full request / response logging
- ✓Cost tracking & attribution
- ✓PII detection (5 types)
- ✓YAML policy engine
- ✓Tamper-proof audit trail (SHA-256 Merkle chain)
- ✓Budget enforcement (daily/monthly limits)
- ✓Session & trace tracking
- ✓18-page real-time dashboard
- ✓Docker + Kubernetes deployment
- ✓Community support
Scout v2
Managed cloud. Compliance automation. Multi-provider routing. Join the waitlist for early access.
- Everything in v1, plus:
- ✓Multi-provider routing (Bedrock, Azure, Vertex)
- ✓SSO / OIDC + role-based access control
- ✓Real-time PII masking (redact before upstream)
- ✓Webhook connectors (Datadog, Splunk, PagerDuty)
- ✓One-click compliance reports (SOC 2, EU AI Act, HIPAA)
- ✓A2A identity & mTLS agent certificates
- ✓Managed cloud with SLA
- ✓On-prem / VPC deployment option
- ✓Dedicated support
Start with v1 today. Get early access to v2 when it launches.
COMPLIANCE DEADLINE
The EU AI Act enforcement deadline is real.
August 2, 2026. Up to €35M or 3% of global revenue for non-compliant AI systems. Scout gives you a tamper-proof audit trail today — for free.
EU AI Act
Up to €35M in penalties or 3% of global revenue for non-compliant AI systems.
SOC 2
Audit trails and access logs are mandatory for enterprise AI deployments.
HIPAA
PHI in AI agent outputs requires complete, tamper-proof audit trails.
FAQ
Frequently asked questions
WHY THIS MATTERS
Real problems. Real stakes.
Wewokeuptoa$47KbillbecauseoneagentgotstuckinaloopcallingGPT-4allnight.Noalertfired.Nobodyknewuntilfinancepingedus.
Based on real incidents reported across AI engineering teams
Govrix tracks per-agent, per-model cost in real time. Set a budget cap — when it hits, the agent stops. No more surprise invoices.
OpenAI's own documentation warns about runaway token usage. Federico Viticci reported burning 180M+ tokens ($3,600/month) using Claude with Apple Shortcuts — with zero cost controls in place.
Built for teams in —
Your agents are bleeding money right now.
See where — in 2 minutes.
Free, open-source proxy. MIT licensed. Self-hosted in 2 minutes. No credit card.