On this page
How Verosek works:
a call, a gateway URL and a key.
Verosek starts with a call. We set it up with you, and you get a gateway URL and a first key. On the Enterprise plan we deploy it with you in your environment. Then two values change in your code: the SDK's base URL and its API key. Your OpenAI, Anthropic or Gemini calls keep their shape, and an MCP client points at the gateway's MCP endpoint with a Verosek key. From the next call, that key's rules decide what runs, and each request leaves a sealed record.
How is Verosek delivered?
You go through five steps, and only the third touches your code.
- [ 01 ]
A call
We talk through where it should run, which model providers you use and which tools your agents call.
- [ 02 ]
A URL and a key
We set it up with you, and you get a gateway URL and a first key. On the Enterprise plan we deploy it with you in your environment.
- [ 03 ]
Two values in your code
Your SDK's base URL points at the gateway, and a Verosek key becomes its API key. Your calls keep their shape.
- [ 04 ]
Rules on each key
Each key carries its allowed models, budget, rate limits, expiry and per-tool rules. Watch mode shows what a rule would have blocked before it enforces. Shield stays off until you enable it and pick a profile for the key.
- [ 05 ]
A decision and a record
A tool call the rules refuse never reaches your system. Each request leaves a sealed trace, scrubbed of secrets and personal data.
What changes in your code?
Two values change: the base URL and the API key. Your model names, messages and tools stay as they are.
OpenAI SDK
from openai import OpenAI
client = OpenAI(
base_url="https://your-gateway/v1",
api_key="vsk_...",
)Anthropic SDK
from anthropic import Anthropic
client = Anthropic(
base_url="https://your-gateway",
api_key="vsk_...",
)Gemini SDK
from google import genai
client = genai.Client(
http_options={"base_url": "https://your-gateway"},
api_key="vsk_...",
)An MCP client
POST https://your-gateway/mcp Authorization: Bearer vsk_...
What do you see afterwards?
You see each decision as it happens, and you can check the record later.
A refusal in your SDK's shape
A tool call the request did not offer, or the key's rules refuse, never runs. The error comes back in your own SDK's shape.
Calls on your deployments
Each call goes to one of your organisation's own deployments, and chat calls fall back in priority order when one fails.
A refusal when something is down
By default a blocking rule that cannot run answers 503 rather than passing, and a response header names any text let through unread. By default a request the trail cannot record is refused before anything runs.
A record you can check
Every request becomes a trace of typed steps, scrubbed of secrets and personal data, then sealed. You can verify a trace through the API. On a self-hosted deployment, an auditor you give the verification key can check it offline, and that key can also make a seal. For anyone else, decisions go to a signed log a third party can check with public keys.
Frequently Asked Questions
Book a call. After it you have a gateway URL and a first key, and you change two values in the SDK you already use: the base URL and the API key.
No. Verosek is not downloadable. It starts with a call: bring one real agent. We set it up with you, you get a gateway URL and a first key, and your application changes two values. On the Enterprise plan we deploy it with you in your environment.
Point the agent's MCP client at the gateway's MCP endpoint with a Verosek key. The endpoint lists the tools that key may use and runs them under the same rules and the same record.
OpenAI, Anthropic, Google, Azure, Ollama and any OpenAI-compatible server. Each call goes to one of your organisation's own deployments.
On the Enterprise plan it can run wholly in your environment, deployed with us: private cloud, VPC or air-gapped. It has no connector for private networks, so it has to reach each server directly.
An app that runs its own tools leaves by changing two values. Tools Verosek ran for you go back to your own code or MCP client.
This page and the product pages are what we publish, and agents can read each page as markdown. Rule details and install steps come with the deployment.
Team is $999 a month, billed yearly. Enterprise has custom terms. Both plans start with a call. See the plans.

