“There is no connector for that system.” We build the custom MCP connector.
Ready-made connectors cover the well-known tools. They do not reach a system on your private network, behind a VPN, or the one your company built years ago. We build that connector. It runs on your infrastructure and the source code is yours.
- For
- Companies whose key data lives in a system with no connector
- You get
- A tested connector, deployed on your infrastructure, with source code and documentation
- Starts with
- A feasibility check, before you pay for a build
- Price
- Part of the project quote
- Your systems
- Stay as they are. Nothing is replaced or migrated
A custom connector is a small service that lets an AI agent read from one of your systems, and act on it, through the Model Context Protocol (MCP), an open standard for connecting AI assistants to external tools. Deep Blue Coding builds connectors for in-house applications, legacy databases and tools that have no public connector, runs them on the client's own infrastructure, and hands over the source code.
The system that matters most is often the one left out
The data people ask about most tends to sit in the oldest system: the warehouse application written in 2011, the pricing database only two people understand, the planning tool that never had an API.
Without a connector to it, every answer has a hole in it, and someone still has to export that part by hand.
What you get
- Feasibility check
- We look at how your team gets data out of the system today and tell you whether a connector is feasible, before you pay for one.
- The connector
- A set of actions agreed with your team: search, read, create, update. Nothing it was not asked to do.
- Per-person sign-in
- Where the system supports it, each person uses their own account and sees through the connector only what they could already see. Where it does not, we agree the limits with you first.
- Deployment on your infrastructure
- It runs next to the system it talks to. The agent reaches it through one secured address that your IT team controls.
- Tests and documentation
- Automated tests, a description of every action and a runbook for your IT team.
- Source code
- Yours, in your repository, with nothing more to pay us for using it.
One we built and run in production
- System
- Microsoft Planner
- Actions
- List, create, update and close tasks, from the chat
- Sign-in
- Each person's own Microsoft account, so they see only the plans they already have access to
- Status
- In production
How it runs
Feasibility
How does data leave the system today: an API, a database, a file export, a screen? We answer with a yes, a no, or a yes with conditions.
Design the actions
With the people who use the system, we list what the agent should be able to look up and do, and what it must never do.
Build and test
Against a copy or a test environment, with real past cases.
Security review
Your IT team reviews the code, the access it needs and where it runs.
Deploy and connect
On your infrastructure, then switched on for the people who need it.
Handover
Code, tests, documentation and a walkthrough for whoever will maintain it.
What we need from you
- Someone who knows the system well, and someone who can grant access to it.
- A test environment or a copy of the data.
- A place to run the connector on your infrastructure.
What it costs
Connectors are part of the project quote, set after the feasibility check. The cost depends on how the system can be reached and how many actions the connector needs. See engagement models.
Questions buyers ask
What is an MCP connector?
It is a small service that exposes one system to an AI agent through the Model Context Protocol (MCP), an open standard for connecting AI assistants to tools and data, created by Anthropic and now governed by a Linux Foundation body. The connector defines the actions available, such as search, read or create, and the agent uses them when a question needs that system.
What if the system has no API?
Then we look at the other ways in: its database, scheduled exports, or files it already produces. If none of them is reliable, we tell you so in the feasibility check and you do not pay for a build.
Where does the connector run?
On your infrastructure, next to the system it connects. The agent calls it through one secured address that your IT team controls. Your data does not pass through servers owned by Deep Blue Coding.
Who owns the code?
You do. The source code goes into your repository with its tests and documentation, and you pay us nothing more for using it.
Can a connector change data, or only read it?
Both, and you choose. Many start read-only. When a connector can create or update records, the actions that matter can be made to wait for a team member's approval. See process automation.
Read next
Last updated 18 November 2026.
Start with one process.
The report your team dreads building, or the approval that sits in an inbox for days. Tell us about it and a senior engineer tells you whether it can be automated and what it would take.
Email social@deepbluecoding.com
We reply within one working day.
Deep Blue Coding