The Model Context Protocol (MCP) is an open standard for connecting AI models to external data sources and tools. Anthropic, the makers of the Claude AI models, introduced it, and it has since been adopted across the AI ecosystem, including by OpenAI, Google, and Microsoft, making it the de facto standard for AI tool integration. Much like USB-C provides a standardized way to connect devices to peripherals and accessories, MCP provides a standardized way to connect AI models to the data and tools they need.
Blues ships two official, hosted MCP servers that work with any MCP-capable AI client:
- Blues Expert helps you build with Notecard and Notehub, with current documentation, request validation, and firmware guidance.
- Notehub MCP lets you ask questions of the data your deployed devices already send, through Notehub IQ.
This guide explains how MCP works and how to start using both servers in your own connected projects.
Understanding How It Works
There are a couple of important concepts to understand when it comes to how MCP works to connect AI models. Here's a quick overview of the major elements in the architecture:
- Hosts, such as Claude Desktop, Cursor or VS Code Copilot, are the AI applications you use.
- Clients, these exist within a host and maintain a 1:1 connection with MCP servers.
- Servers, these provide context, tools, and prompts to clients.
- Transport Layers, this is how a client communicates with a server. MCP defines two standard transports:
stdiofor local servers that run as a subprocess, and Streamable HTTP for remote servers (this replaced the older, now-deprecated HTTP+SSE transport). Both Blues servers are remote HTTP servers.

Specifically within an MCP server, there are some key components:
- Tools: These are the actions that the MCP server can perform. This could be querying an API, updating a database, or controlling a hardware device.
- Resources: These are the data sources that the MCP server can share with the client, such as API docs, usage examples, etc.
- Prompts: These are reusable prompt templates that a server offers to the client, such as "Analyze a codebase" or "Debug a project, following a specific set of instructions", which a user can pick instead of writing the instructions from scratch.
The model, working through its client, decides which tool or resource it needs and requests it from the MCP server. The server runs the tool or returns the resource, and the model uses the result to answer you or take its next step.
How Can MCP Improve Your Development Process?
Most code editors and AI clients, including Claude, Cursor, VS Code, and Codex, now support MCP servers, so adding one changes what your existing assistant can do.
One Integration for Every Client
A service that offers an MCP server works with every MCP-capable client, so you don't need a separate plugin for each editor or chat app. You add the server once per client, usually by pasting a URL, and the same tools are available wherever you work.
Current Information Instead of Training Data
A model's knowledge stops at its training cutoff, so it can confidently describe an API that has since changed. An MCP server can give it current documentation and data on demand, so the model can check an API as it works instead of relying on memory.
Actions and Answers from Real Systems
Without tools, a model can only tell you what to do. With an MCP server, it can query a service's API, read live data, and (with your approval) make changes. Rather than learning the nuances of Notehub's APIs yourself, for example, you can ask the question in plain language and let the model call the right tools.
Access You Control
Remote MCP servers that work with your own data typically use OAuth, so you sign in to the service and choose what the AI client may access. You can limit it to read-only access or a single project, and revoke it at any time from the service's side.
Getting Started with Blues Expert
The fastest way to put MCP to work while you're building with Blues is the Blues Expert MCP server. It's a hosted server, so there's nothing to clone, build, or run locally, and it gives your AI client:
- Notecard API documentation and request validation
- Search over the latest Blues documentation
- Expert analysis from LLM specialists with deep Notecard, Notehub, and IoT knowledge
- Firmware development guidance for Arduino, C, Zephyr, and Python
You connect to it with a single URL: https://mcp.blues.io/expert/mcp.
For example, to add it to Claude Code:
claude mcp add --transport http blues-expert https://mcp.blues.io/expert/mcpBlues Expert also works with Claude Desktop, GitHub Copilot (in VS Code and the Copilot coding agent), Cursor, and most other MCP clients. See the Blues Expert MCP documentation for client-by-client setup instructions.
Once connected, you can ask things like:
- "What's the latest firmware for my Notecard, and how do I update it?"
- "Validate this Notecard request and explain what each field does."
- "Show me how to configure a Notecard to send a note when it detects movement."
Asking Questions of Your Fleet with Notehub MCP
Once devices are deployed, the questions change from "how do I build this?" to "what are my devices doing?". That's what Notehub MCP is for. It's the interface to Notehub IQ, the intelligence layer for Notehub: it turns plain-language questions into SQL queries against the Notehub data you already collect, runs them, and reasons over the results.
Notehub MCP is also a hosted HTTP server, at https://mcp.notehub.io/mcp.
To add it to Claude Code:
claude mcp add --transport http notehub https://mcp.notehub.io/mcpIn Claude Desktop or on the web, you can also search for "Notehub" in Claude's Connectors directory.
Because Notehub MCP works with your own product data, the first connection opens a browser window where you sign in to Notehub and choose what your agent can access.
You can give it access to all of your projects or just one.
Pick a My Privileges option the first time you connect, because the next step, onboarding a project and creating its datasets, makes changes to the project (and requires the Owner or Developer role on it).
Onboard a project and create datasets from the Notefiles you care about, including System Notefiles like _session.qo and _health.qo if you want to ask about signal strength, data usage, or power.
Once your projects are onboarded, you can reauthorize with a Viewer Privileges option if you want your agent to stay read-only.
The Notehub MCP documentation walks through each step.
Once your datasets are ready, you can ask things like:
- "Which devices are actually down versus just syncing on an expected periodic basis?"
- "Which devices are burning the most cellular data, and is it payload size or sync frequency driving it?"
- "Rank my battery-powered units by how close they are to brown-out."
Notehub MCP can make changes to your projects, but it always asks for your approval first. And because it's one more tool your agent can reach, you can pair it with your other connectors, such as Slack or a ticketing system, or run a question on a schedule so your agent reports only when something looks wrong.
To try Notehub IQ in your browser before connecting your own agent, see the Using Notehub IQ tutorial.
Which Blues MCP Server Should You Use?
The two work well together: Blues Expert while you write firmware, and Notehub MCP once your devices are in the field.
| Blues Expert | Notehub MCP | |
|---|---|---|
| Use it to | Build with Notecard and Notehub | Ask questions of your deployed devices' data |
| Server URL | https://mcp.blues.io/expert/mcp | https://mcp.notehub.io/mcp |
| Sign-in | None | Your Notehub account, with access you choose |
| Works with | Blues documentation, the Notecard API, and firmware libraries | The datasets you create from your Notehub projects' Notefiles |
| Example question | "Validate this Notecard request." | "Which devices went quiet in the last 24 hours?" |
Conclusion
MCP has quickly become the standard way to connect AI models to real-world data and tools, and it can radically improve your development workflows. With Blues Expert, you can chat with an on-demand Notecard and Notehub expert from whichever MCP-capable client you already use. With Notehub MCP, the same client can answer questions about your fleet from the data your devices already send.
The fastest way to see the value is to try it: add both servers to your AI client, ask Blues Expert to help with your next Notecard project, and ask Notehub MCP what your devices have been up to.

