What this gateway is
REVsign publishes a single address your AI assistant can use to query the engine. Instead of opening the panel, you open the chat you already use and ask for what you need in plain language: what was said about a company this week, which companies in the census belong to a vertical, whether a domain is still alive, whether a tax ID is valid. The assistant translates the request, queries, and answers in the same conversation.
What matters is what does not change when you do. The rules governing the Unified Signal Layer are the same through this gateway as through the panel: the same permissions, the same exclusions, and the same human approval before anything goes out in your company’s name. The assistant is a vehicle, not an exception.
What you can ask for, and which use case it maps to
Read queries and logging a signal go out through this gateway. Each one is the conversational version of a case that already exists in the engine, and you can see them all in Use Cases.
| What you ask the assistant | Use case |
|---|---|
| A company’s recent signals, or the full detail of one signal with its quote and its signature | Market Radar & Competitive Intelligence |
| The morning review, what came up since yesterday and deserves a decision | GTM drafts for asynchronous action |
| Which companies in the census belong to a vertical, a country or a band, with their verification status | Census and vertical segmentation in Argentina |
| An account’s full context: its history, its corporate relationships and who decides the purchase | Co-Thinker: auditable institutional memory, and Corporate relationship graph |
| Whether a domain resolves, redirects or is abandoned, before spending on enriching it | Domain hygiene and verification (Gate 1) |
| Whether a tax ID is valid according to the arithmetic that defines the number itself | Census and vertical segmentation in Argentina |
| What changed in the software an account uses between two reviews, and how much that change weighs | Changes in the software each account uses |
| Whether a domain is excluded from commercial contact, and for what reason | Transactional dynamic exclusion purge |
| Log a new signal from the conversation | Unified Signal Layer |
You do not need to learn command names. The assistant discovers on its own what it can ask for once connected, so describing what you want to know is enough.
How to connect it
The address is https://mcp.revsignlab.com/mcp, and access is by invitation. With your credential enabled, Claude Code connects in a single line:
claude mcp add --transport http revsign https://mcp.revsignlab.com/mcp \ --header "Authorization: Bearer YOUR_KEY"
From the OpenAI API, the same address is declared as a tool for the model, with your key in the header and approval for each call turned on.
If you do not have a credential yet, request access. Without an enabled user the address returns nothing, apart from the open query for a signal already published, the same one anyone can read on the public radar.
The limits, in writing
The gateway is deliberately narrower than the engine, and that is not a setting anyone can loosen from outside.
Contact data does not come out through here. Looking up a person and seeing which product you have not yet sold to an account are out by construction. There is no option that enables them.
Writing still goes through a person. A signal logged from the chat is held on arrival, exactly as in the panel. It is a gateway for working with you, not a process that acts on its own.
Anything that spends an external provider’s credit is enabled one at a time, by a written decision, never by default.
Each credential has its own per-minute request cap, so an assistant stuck in a loop cannot drag the engine down or run up the cost.
The whole gateway can be switched off without touching the engine, if it ever becomes necessary.
Frequently asked questions
Which assistants does it work with today?
Claude Code and the OpenAI API, which come in with their own revocable key. Signing in with your account, which is what claude.ai, ChatGPT and Gemini CLI need, is not open yet.
Can the assistant send an email or publish something in my company’s name?
No. No outward action leaves through this gateway. Anything logged from the chat is held awaiting human approval, under the same standard that governs the rest of the engine.
If I switch assistants, do I have to redo the integration?
No. It is the same address for all of them, and there is no per-client integration to write. Switching assistants means pasting the address somewhere else.
$ curl revsign.os/docs/connect-your-ai-assistant
→ {
title: "Connect the engine to your AI assistant",
answers: "How do I connect my AI assistant to my company data without writing an integration?",
schema: [TechArticle, FAQPage]
}