MCP Server?

As mentioned in previous posts, our latest rebuild was built in part to fix some annoying design decisions we had made, but it was mostly done to position QC Database for the age of AI. This involves adding some optional cloud AI tools within QC Database, and more importantly, adding MCP Server support.

"MCP" stands for Model Context Protocol, which Anthropic announced as a new standard in 2024. Think of this protocol as a standardized way for AI to talk to enterprise software (like QC Database) or in Anthropic's words, "a new standard for connecting AI assistants to the systems where data lives."

flowchart LR
    subgraph Host["AI Host Application (e.g. Claude)"]
        LLM[LLM]
        Client1[MCP Client 1]
        Client2[MCP Client 2]
        LLM <--> Client1
        LLM <--> Client2
    end

    subgraph Servers["MCP Servers"]
        Server1[MCP Server A]
        Server2[MCP Server B]
    end

    Client1 <-->|JSON-RPC over stdio / HTTP| Server1
    Client2 <-->|JSON-RPC over stdio / HTTP| Server2

    subgraph Capabilities["Server Capabilities"]
        Tools[Tools<br/>callable functions]
        Resources[Resources<br/>readable data/context]
        Prompts[Prompts<br/>reusable templates]
    end

    Server1 --> Tools
    Server1 --> Resources
    Server1 --> Prompts

    subgraph External["External Systems"]
        DB[(Databases)]
        API[Web APIs]
        Files[Local Files]
    end

    Server2 --> DB
    Server2 --> API
    Server2 --> Files

MCP Servers took off quickly in the software development world. In 2026, it is taking the broader business world by storm.

You can read more about the standard here. MCP is now an open standard, similar to HTTP. In December 2025, Anthropic handed control of the standard to the newly created AAIF.

QC Database MCP

image We have redesigned our API with a focus on MCP compatibility. Our pre-built MCP server acts as a USB cable between your AI agent, and QC Database. This allows you to build software integrations on the fly or generate custom reporting at a moments notice. You can even add items, update items, or delete items (all deleted records are preserved) on QC Database by just telling your AI Agent to perform these actions for you. Any actions performed by your AI agent are done in your name, using your credentials, so configure your permissions carefully. We recommend setting read-only endpoints to run without approval, all edit and delete endpoints to require approval, and create endpoints to either require approval, or run without approval, depending on the data type.

MCP Servers are simple applications, usually written in python or javascript. They perform simple bridging tasks between AI Agents and API's. We open-sourced our MCP server so you can customize it as needed. It is released under the very permissive MIT license so you can modify it, run it locally, host it on your network, or host it on the wider web with your company's processes baked-in.

Hosted MCP Server, with instructions to connect: MCP.QCDatabase.AI

Open Source Git Repo for developers. the standard QC Database MCP server

Install with Git

git clone https://github.com/ProcessQualitySolutions/qcdatabase_mcp.git

In this blog post, I will post screenshots from some basic operations. The BIG DEMO is coming next week.

Mini Demo 1 - Asking basic questions

Claude and Chat GPT know a lot, but they don't know how many Flange Torque reports you have on your job. You could log into QC Database, navigate to the Custom Forms page, and then look at the count at the top of the page. That is about 5 or 6 clicks. With our new Remote MCP Server, you can just ask... image

P.S. Notice that I used voice-to-text (mic icon). AI can really contribute to transparency on a project, and this is only scratching the surface.

Mini Demo 2 - Updating a List

A very simple idea, and a great use case for AI. If you are a CWI testing welder John Doe, you can add him to the welder list by talking into your phone. You can do this because the MCP server is connected using your account, and your permissions restrict what it can do. Claude can only add a welder if you are allowed to by the project admin. You will rarely see hallucinations on such a simple task. In fact, AI can often prevent YOU from making a mistake...

Let me try to add a welder that we already have on the list (oops)... image

OK, let's actually add a welder... image

At the end we get this: Our welder list is updated and we avoided a small human error. image

Mini Demo 3 - Form Auto-Completion

Mobile-friendly digital forms have made paperwork a lot easier in construction, but let's take it a step further. Let's tell Claude to initialize a Receiving and Inspection Report for a sheet metal hauling truck that I know is coming later, even though I don't know what it is hauling. First, because I started a new chat, Claude is making sure I am still on the same project. image

Claude also asked me what I meant when I said "Today" because I asked it to create the form just after midnight. It asked about all the required fields that I didn't have info for, and recommended writing "TBD" in those fields. It even suggested using a field note instead of a partially filled form as a placeholder (probably a better workflow). Eventually, It finally created my new form and gave me a link. Here is a screenshot of the PDF. image

Claude gave me a lot of hassle on creating that RIR, but he eventually created it. I think this highlights an interesting role that AI can play in construction: the "Company Procedure Enforcer". Claude was trying to get me to use QC Database notes for internal reminders instead of a sloppy, half-filled form. If I told Claude that 'TBD' forms are our company procedure, he might have done it more quickly. This is why we have open-sourced our MCP server. Currently, the MCP server facilitates QC-Database's functions and workflows, but a custom QC Database MCP Server could act as a framework for your company's procedures. Claude, or other AI agents, will act as the enforcers. It sounds annoying, but the AI agent is willing, and now able, to do the paperwork for you. image

Stay tuned for the BIG DEMO!

The next blog post will include a video and a serious demo of AI-driven processes that will save your company real time, effort, and money. If you want a sneak peek, take a look at our new SKILLS page. These skills are also open-sourced, work well with the QCDatabase MCP server, and can be customized to your needs and your company's processes. At QC Database, are betting the house on AI (in terms of the design of QC Database, not investment advice), and next week, you will see why.