MCP BridgeRequest service
For small teams who want Claude or ChatGPT to answer from their own data

Ask Claude or ChatGPT about your own data, safely

Both Claude and ChatGPT can now use custom connectors built on the Model Context Protocol (MCP). MCP Bridge builds one for you: we connect your Google Sheet, Airtable base, Postgres or MySQL database, or internal API, decide with you exactly which data and actions the AI may use, host it securely, and add it to your team's Claude or ChatGPT. Paid once for the build; hosting is included for a year.

Request service See pricing
  • A remote MCP server with sign-in, so only your team's accounts can use it, added as a custom connector in Claude and ChatGPT.
  • Read-only by default: you choose which tables, columns and actions are exposed, and every request is logged.
  • Tested with the questions your team actually asks, like 'which invoices are overdue?' or 'what did we quote this customer?'.
No card needed to requestDelivered within 2 business daysPay once, no subscription
Hosted and monitoredWe host and monitor it for a year, with a log of every…
Claude

Claude lets users add a custom connector by pasting a remote MCP server URL under Settings, Connectors, on its paid plans.

Source: lagrowthmachine.com
ChatGPT

ChatGPT supports custom MCP connectors through Developer Mode; the server must be publicly reachable, and business plans may need an admin to enable it.

Source: zuplo.com
Why now

AI assistants are useful until they need your data

Claude and ChatGPT are good at answering questions, drafting and summarizing, but they don't know your stock levels, your customer list or what was quoted last month. Copying data into a chat each time is slow, risky and out of date the moment you paste it.

The Model Context Protocol fixes this: a small server describes your data and actions, and the assistant calls it when it needs to. Claude supports custom remote connectors on paid plans, and ChatGPT supports custom MCP connectors through its developer mode. What's missing for most small teams is someone to build the server, secure it and keep it running.

MCP Bridge does that as a one-time job: a connector scoped to exactly what you allow, hosted for you, and tested with real questions.

Without it

  • Official connectors: Many SaaS tools now ship their own MCP connectors; if yours does, use it, and we'll tell you. MCP Bridge is for…
  • Building it yourself: The MCP SDKs are open source and a developer can build a basic server in a day. Sign-in, scoping, hosting and logs are…
  • Copying data into chats: Free and quick for one-off questions. It leaks data into chat histories, goes stale immediately and doesn't scale…

With MCP Bridge

  • A remote MCP server with sign-in, so only your team's accounts can use it, added as a custom connector in Claude and ChatGPT.
  • Read-only by default: you choose which tables, columns and actions are exposed, and every request is logged.
  • Tested with the questions your team actually asks, like 'which invoices are overdue?' or 'what did we quote this customer?'.
  • Your data sources
  • Read-only by default
How it works

How it works

  1. 1

    Pick the data

    On a 30-minute call we agree the source (Sheet, Airtable, database, API), which tables or endpoints, and the ten questions your team wants answered.

  2. 2

    We build the connector

    A remote MCP server with read-only tools by default, sign-in for your team, and rate limits, built in about 3 business days.

  3. 3

    Add it to your assistant

    We add the connector to your Claude workspace and ChatGPT, or walk your admin through it, and test the ten questions together.

  4. 4

    Hosted and monitored

    We host and monitor it for a year, with a log of every request. Changes to the data or tools are a small fixed fee.

What it does

Features

Your data sources

Google Sheets, Airtable, Notion databases, Postgres, MySQL, SQL Server, or any internal REST API with a key.

Read-only by default

The assistant can look things up but not change them. Write actions, like adding a note or creating a draft order, are added only if you ask, with a confirmation step.

Column-level control

You choose which tables and columns are visible. Salaries, card numbers or personal notes can be left out entirely.

Sign-in for your team

OAuth sign-in so only your team's accounts can use the connector; links shared outside the team do nothing.

Works in Claude and ChatGPT

Built to the open MCP standard, so the same connector works in Claude, ChatGPT and other MCP clients your team adopts later.

Request log

Every call the assistant makes is logged with the user, the question's tool call and the rows returned, so you can audit what was read.

A year of hosting included

Hosting, monitoring and security updates for 12 months, then $9 a month if you want us to keep running it, or we hand over the code.

A look inside

A look inside

What you get: the working view and the delivery report.

The working viewA remote MCP server with read-only tools by default, sign-in for your team, and rate limits, built in about 3 business days.
Your delivery reportWe agree the scope within 2 business days of your request and deliver most connectors within a week.
Built for

Who it's for

Small businesses on spreadsheets

Stock, orders, customers or bookings in Google Sheets or Airtable, and a team that already uses ChatGPT or Claude.

Teams with an internal database

A Postgres or MySQL database behind an internal tool, and people who keep asking the one developer to run queries.

Agencies and consultants

Give each client an assistant that knows their data, without building MCP servers yourself.

SaaS founders

Offer your customers an MCP connector for your product without pulling your own team off the roadmap.

Works with

Plans

Pricing

One-time build price, paid once after we agree the scope. Hosting and monitoring for the first 12 months are included; after that $9 a month to keep it running, or we hand over the code. Founding customer pricing applies to early-access orders.

Sheet

$99 one-time

One spreadsheet or Airtable base

  • 1 Google Sheet or Airtable base
  • Up to 5 read-only tools
  • Sign-in for up to 5 users
  • Added to Claude and ChatGPT
  • 12 months hosting
Request service
Recommended

Database

$249 one-time

A database or internal API, scoped carefully

  • 1 database or REST API
  • Up to 15 tools, read-only by default
  • Column-level access control
  • Sign-in for up to 25 users
  • Request log
  • 12 months hosting
Request service

Multi-source

$590 one-time

Several sources and write actions

  • Up to 3 sources
  • Write actions with confirmation
  • Unlimited users in your workspace
  • Everything in Database
  • Code handover on request
  • Priority support
Request service

Compared with the alternatives

Official connectors
Many SaaS tools now ship their own MCP connectors; if yours does, use it, and we'll tell you. MCP Bridge is for spreadsheets, databases and internal tools that don't have one.
Building it yourself
The MCP SDKs are open source and a developer can build a basic server in a day. Sign-in, scoping, hosting and logs are the parts that take longer and matter most.
Copying data into chats
Free and quick for one-off questions. It leaks data into chat histories, goes stale immediately and doesn't scale beyond one person.

What changed

  • Claude lets users add a custom connector by pasting a remote MCP server URL under Settings, Connectors, on its paid plans. Source: lagrowthmachine.com
  • ChatGPT supports custom MCP connectors through Developer Mode; the server must be publicly reachable, and business plans may need an admin to enable it. Source: zuplo.com
Getting started

From call to connector

We agree the scope within 2 business days of your request and deliver most connectors within a week.

  1. Day 1: Request the service; we book a 30-minute scoping call.
  2. Day 2: Scope agreed: sources, tools, columns and ten test questions.
  3. Day 3–5: Connector built, secured and deployed.
  4. Day 6–7: Added to Claude and ChatGPT and tested with your team.

Security and data

  • Connectors are read-only by default, and every tool and column exposed is agreed with you in writing.
  • OAuth sign-in restricts use to your team's accounts; credentials to your data sources are encrypted and never shown to the assistant.
  • Every request is logged with the user and the tool called; logs are kept 90 days or turned off on request.
  • We access your data sources only through the scoped credentials you create for the connector, which you can revoke at any time.

Questions

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants call tools and read data through a small server. Claude and ChatGPT both support custom MCP connectors, so one server works in both.

Which Claude and ChatGPT plans support custom connectors?

Claude supports custom connectors on its paid plans, and ChatGPT through Developer Mode on paid plans, with admin approval on business workspaces. We check your plan on the first call.

Can the AI change or delete our data?

Not unless you ask for write actions. Connectors are read-only by default; write actions are added one by one, with a confirmation step in the assistant.

Where is our data stored?

It stays in your sheet, database or API. The connector reads it on request and doesn't keep copies, apart from the request log you can turn off.

What happens after the first year?

You can keep hosting with us for $9 a month, or we hand over the code and deployment instructions so you can run it yourself.

Can you build one for our product, for our customers?

Yes. The Multi-source plan covers a customer-facing connector for a SaaS product; larger scopes are quoted after a call.

Ready when you are

Tell us about your setup. We confirm by email and get you running within 2 business days.

Request service