Dex MCP Server: Supercharge Your AI Agents with Contact Relationship Management

If you're building AI agents that need to interact with contacts, notes, and reminders, you've probably hit a wall when it comes to giving them real CRM capabilities. The Dex MCP Server solves exactly that problem—it's a Model Context Protocol server that plugs your AI agents directly into the Dex API, letting them create contacts, manage notes, and set reminders just like a human would.

Think of it as giving your AI assistant a proper contact book instead of a messy pile of sticky notes. Let's walk through what makes this tool worth your time.

What Problems Does Dex MCP Server Solve?

Most AI agents are great at conversation but terrible at remembering people. They can't track who you talked to last week, what you discussed, or when you need to follow up. The Dex MCP Server changes that by giving your agents three core superpowers:

Contact management that actually works. Your agents can create, search, and update contacts with full details—names, companies, job titles, emails, phone numbers, the whole package. No more asking "wait, which John are we talking about?"

Notes that stick to people. Every conversation, meeting, or interaction can be saved as a note attached to the right contact. Later, your agent can search through all those notes to find that thing someone mentioned three months ago.

Reminders that don't get lost. Set follow-up reminders linked to specific contacts, with support for recurring tasks. Your agent can actually nudge you when it's time to reconnect with someone.

👉 Manage your professional network effortlessly with Dex's intelligent contact management

Setting Up the Server (It's Easier Than You Think)

You'll need Node.js 18 or later and a Dex API account. The setup process is straightforward:

Clone the repository and install dependencies with standard npm commands. The key step is configuring your Dex API key—and here's where the developers did something smart. Instead of hardcoding keys (which always ends badly), you create a .env file based on the included .env.example. Pop your API key in there, and the server loads it automatically. The .env file is already gitignored, so you won't accidentally commit your secrets to GitHub.

Build the project with npm run build and you're ready to roll. If you're using AnythingLLM, add the server to your MCP configuration by pointing it at the compiled dist/index.js file.

The Full Toolkit: 18 Tools Across Three Categories

The server gives you 18 different tools split across contacts, notes, and reminders. Here's the breakdown:

Six contact tools let you get, search, create, update, and delete contacts. Pagination support means you can handle massive contact lists without choking your system.

Six note tools cover everything from fetching all notes for a specific contact to searching across every note you've ever written. Each note is stored as a timeline item in Dex, which means they show up chronologically in your contact's history.

Seven reminder tools (yes, seven—there's an extra one for marking reminders complete) give you full lifecycle management. Create reminders with due dates, set them to recur, search by content, and mark them done when you follow up.

One nice touch: when creating notes, you can specify meeting types using either display names ("Coffee") or API enum values ("coffee"). The system automatically maps between them, so you don't need to memorize API conventions.

👉 Keep track of every interaction and never miss a follow-up with Dex

Real-World Scenarios Where This Shines

Imagine you're building an AI sales assistant. It can automatically create contacts from email introductions, add notes after each call, and set reminders for follow-ups. When you ask "who did I talk to about blockchain projects last quarter?", it searches through all your notes and gives you names.

Or say you're managing a network of freelancers. Your agent tracks when you last worked with each person, what projects they handled, and when to check in about new opportunities. It reminds you when someone's birthday rolls around or when it's been too long since you connected.

The key is that your agents aren't just chat interfaces anymore—they're actively managing relationships on your behalf.

Built Right: TypeScript, Validation, and CI

Under the hood, the server uses TypeScript for type safety (so fewer "undefined is not a function" nightmares), Zod for schema validation (your data stays clean), and Axios for HTTP requests to the Dex GraphQL endpoint.

The project includes ESLint configuration and GitHub Actions CI that runs type checks and linting on every pull request. There are issue templates for bug reports and feature requests, plus a PR template to keep contributions consistent.

Security is taken seriously—the repository includes a git history scrubbing script in case secrets ever get committed accidentally, though the .env setup makes that unlikely.

Who Should Use This?

This server is perfect if you're building AI agents that need to remember people and relationships. Sales automation, customer success tools, personal CRM assistants, networking apps—anything where contacts and follow-ups matter.

It's also great for developers who want to experiment with the Model Context Protocol without building everything from scratch. The code is clean, well-documented, and MIT licensed.

Getting Started

The repository includes comprehensive documentation for both setup and usage. If you run into issues, there are clear escalation paths: check Dex docs for API questions, the MCP specification for protocol questions, and open a GitHub issue for server-specific problems.

For development, there are npm scripts for watch mode compilation, ts-node for quick iterations, and all the usual build tooling you'd expect from a modern TypeScript project.

The bottom line: if your AI agents need to manage contacts, take notes, and set reminders like a real assistant would, the Dex MCP Server gives you all the tools to make that happen—without reinventing the wheel or wrestling with API documentation for hours.