Verosek
Book a call

[ Compliance First ]

One Layer between your Apps and every AI Agent

Route LLM calls to any provider. Execute MCP tools with per-key access control. Scan every request, response, and tool output for injection, PII, and off-topic drift. Sign and audit every step.

Book a callRead documentation

What runs unguarded in every AI stack

Routing and scanning are solved problems. These five aren't — until you add a governance layer.

[ Threat 01 ]

Pre-execution policy verdicts

Every request gets a verdict before anything runs. ALLOW, BLOCK, MODIFY, or HOLD — decided deterministically, not by a model. Nothing executes until the policy says so.

[ Threat 02 ]

Hallucination grounding via a local judge

A judge model runs entirely inside your infrastructure and scores every response against the context it was supposed to be grounded in. Ungrounded claims are flagged or blocked, depending on your profile.

[ Threat 03 ]

Cumulative session monitoring

Risk is tracked across an entire session, not just single requests — catching slow, multi-step attacks that per-request scans miss.

[ Threat 04 ]

Policy as Code

Guardrails are defined as versioned, testable code — reviewed, diffed, and shipped through the same pipeline as the rest of your stack.

[ Threat 05 ]

Tamper-evident hash chain

Every decision is linked into an append-only hash chain, so any attempt to alter the audit trail is immediately detectable.

How it works

Four components. One binary. No integration gap.

Each component handles a distinct job. Together they share identity, the audit trace, and ship as one.

[ Layer #01 ]
import openai

client = openai.OpenAI(
    api_key="vsk_prod_04",        # Verosek virtual key
    base_url="https://gateway.your-domain/v1",  # ← one line changed
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Summarise Q3 results"}],
)
# Policy enforced. Every step audited.
print(response.choices[0].message.content)

AI Gateway

Route any SDK to any provider. Twelve OpenAI endpoints plus native Anthropic and Gemini. Change one line — swap base_url and key. Cross-SDK translation runs on the fly. Weighted fallback, priority routing, 3-fail cooldown. Under 30ms P99 overhead.

15 endpoints · 3 SDKs · <30ms P99

[ Layer #02 ]
IDCHECKVERDICTLATENCY
CHK-013Prompt injectionBLOCK4ms
CHK-014Jailbreak attemptPASS4ms
CHK-015PII in inputMODIFY6ms
CHK-016Secrets in inputPASS3ms
CHK-017Toxicity in responsePASS5ms
CHK-018PII in responsePASS4ms
CHK-020Indirect injectionPASS8ms
CHK-022Session driftWARN2ms
8 checks run:
1 blocked
1 modified
1 warning
36ms total

Shield

24 security checks across input, output, tool output, and session drift. Deterministic where possible. Offline ML everywhere — nothing about your prompts leaves your network. Four profiles: none, baseline, strict, custom. Graduate from shadow mode when false-positive rate hits zero.

24 checks · 6 PII languages · 4 profiles

[ Layer #03 ]
Agent Request
query_payments → "show all payments"
Policy · 4 Rules Applied
READAllowed
Writeblocked
limitadded
card_numberhidden

MCP Tools

50+ tool connectors across databases, SDLC, knowledge, search, and observability. Per-tool, per-key access control. Argument rewriting and PII redaction recorded as MODIFY in the trace. Credentials encrypted with AES-256-GCM at rest.

50+ connectors · 6 access rule types · AES-256-GCM

[ Layer #04 ]
allow9f3c…a71b
modifyb82e…4c0d
block1d77…e9af
Verifiedchecked with your own key

Audit Vault

HMAC-SHA256 signed receipt for every decision, each one chained to the last. Tamper-evident — change a byte and the chain breaks. SessionTrace per request: ordered steps, aggregate counters, signature. Verify offline with one CLI command, no call back to Verosek. NIST AI RMF and EU AI Act evidence bundles drawn from real trace data.

HMAC-SHA256 · OFFLINE VERIFY · NIST + EU AI ACT

All that with just 2 line of code.

Keep your SDK.
Keep your models.
Keep your code.

import openai
 
client = openai.OpenAI(
    api_key="vsk_...",
    base_url="http://your-gateway/v1",
)

Verosek is a drop-in replacement for the OpenAI, Anthropic, and Gemini SDKs.

Receipts, not claims

Every decision signed. Every trace verifiable offline.

Verosek signs every decision and routes it to your apps; an auditor verifies each receipt offline.

An auditor verifies each receipt with one CLI command — no network call back to Verosek.

Beyond the platform

We don't just claim Shield works, we prove it.

Continuous adversarial testing against every check — every finding ships with hashes you can recompute, not screenshots.

77Attack techniques
34Built for MCP
98%Accuracy
See the red-team results
Testimonials

See what people are saying about Verosek

And we are just getting started...

Indirect injection through tool returns is the gap in agent safety, and most guardrail products don't scan it. Verosek does. The cumulative session-level monitoring also catches the slow distributed version. From a research standpoint, that's the right place to put the work

Sumeet Motwani

Oxford ML PhD, ex-BAIR, alignment researcher

When you're running agents in production, key-level access isn't enough. I want one agent to read a table but not modify it, or hit one endpoint but not the next. Verosek draws that line per tool and per argument. That's the layer most gateways skip.

Sarvagya Kulshrestha

Founder, Backdoor (Ex-Antler, Seekho, OTPless)

Keep buildling, you do great stuff

Dane Knecht

CTO, Cloudflare

Fifteen years in infra, same lesson every time: fail-open is a fiction people end up regretting. Verosek's fail-closed default with an SDK circuit breaker is the right posture for production. If the policy layer is down, the call doesn't land. Full stop.

Rushabh Shah

15+ yrs infra & security leadership (EY Parthenon, Lyft, Dialpad, Cisco, Vox)

Building across multiple AI surfaces, you end up writing wrapper code per provider. Verosek lets the same client hit OpenAI, Anthropic, or Gemini without rewriting anything. One integration point, one audit trail, regardless of which model is running underneath.

Viren Inaniyan

Co-Founder/CEO, TruCommerce (IIT Bombay)

Coming from fintech ops, the compliance posture is what stands out. Mapped reports for NIST AI RMF and EU AI Act, signed traces an outside auditor can verify offline, deployment that keeps data inside your VPC. That's the unglamorous list that decides whether AI agents clear procurement.

Tanish Shah

EIR @ Trueway, ex-PhonePe, ex-Roland Berger (IIM Calcutta, BITS)

The interesting design choice: the policy engine is deterministic and version-controlled. A table of checks you can read, diff, and reason about. Not an LLM judging each call. For something in the request path, that's the only approach that holds up under review.

Pratham Grover

ML, Microsoft Core AI (Ex- Seimens, Samsung)

Pricing

Govern every agent. Prove every decision.

Scan, redact, and log every request across every model and key. No hidden fees, no lock-in, cancel or switch anytime.

[ Team ]
$999/ month, billed yearly.

Overage: $0.50 / 1k requests · $2 / 1k scans

  • 25 virtual keys
  • 5 MCP connections
  • 500k requests / month
  • 500k Shield scans / month
  • All Shield profiles: baseline, strict, custom
  • HMAC-signed receipts
  • 5 Users (admin/member/auditor)
  • Email Support 24hr response
  • 30-day audit retention
Book a call

Frequently Asked Questions

SaaS managed, self-hosted (Docker), or on-prem with a lightweight Connector Agent that keeps your databases inside your VPC — the agent reaches out to Verosek, not the other way around. No inbound firewall changes.

Yes. The self-hosted and on-prem deployments run fully offline — no outbound calls to Verosek are required for the gateway, Shield, or audit signing.

Under 30ms P99 for routing and deterministic checks; offline ML checks add a few milliseconds depending on the profile you enable.

All credentials are encrypted at rest with AES-256-GCM, with keys managed in your own KMS on self-hosted and on-prem deployments.

You do. On self-hosted and on-prem, the audit signing key lives in your own infrastructure — Verosek never holds it.

Yes. Every receipt is HMAC-SHA256 signed and chained; an auditor can verify the full chain offline with a single CLI command, no call back to Verosek.