All resourcesGuidesCoursesTips

5 min read

Build a doc with Coda's MCP

A quick-start guide to building with Coda MCP.

Create pages. Read content. Update tables. Apply formatting. All through natural language prompts.
Coda's MCP (Model Context Protocol) connects your AI client directly to your docs, so your agent doesn’t just chat about work; it can actually do the work. In this guide, you’ll build a quick team task tracker to experience Coda MCP in action. By the end, you’ll have:
  • A working Coda Doc.
  • Structured data to act on.
  • A feel for what Coda's MCP can really do.
Let’s get started.
You'll learn how to:
  • Create a doc with a prompt
  • Structure and refine data
  • Update multiple rows in bulk
What you’ll need:
  • Access to Coda MCP
  • A few copy-and-paste prompts
  • An AI client connected to Coda MCP

Quick terms

 The Coda MCP lets your AI client talk to your doc, so that it can create, read, and update your docs, rather than just providing suggestions. A few terms to know:
  • AI client — The app where you chat with your AI (e.g., ChatGPT, Claude, Cursor, or similar). You share doc links and type your prompts here. The client connects to Coda via MCP.
  • Agent — The AI inside that client. It reads your prompts, figures out what to do, and calls on the Coda MCP to get it done.
  • Coda MCP — Coda's implementation of the open Model Context Protocol standard. It connects your agent to your docs, giving it tools to read, create, and update your docs.
For more on how the Coda MCP works, check out Getting started with the Coda MCP.

How this guide works

 You’ll complete five short steps. For steps with a prompt:
  1. Copy the prompt.
  2. Paste the prompt into your AI client.
  3. If prompted, approve any MCP permissions.
That’s it. Your agent handles the rest. Most AI clients will ask for permission before running MCP tools. When prompted, you can allow all actions, approve specific tools, or review each action individually. You’re in control. Manage or revoke these permissions in your AI client’s settings at any time.

Step 1. Connect Coda MCP

 Before you start building, confirm Coda MCP is connected to your AI client of choice. If you haven’t set it up yet, follow the steps in Connecting to Coda MCP. Setup is usually quick: connect your AI client, approve permissions, and you’re ready to go. If you run into any issues, restarting the connection process does the trick. When you’re connected, come back here.

Step 2. Create your first doc

 With Coda MCP connected, let’s build something real. You describe what you need, and your agent brings it to life. Instead of starting with a blank canvas, you’re going to ask your agent to create a simple starter workflow. You'll create it in your workspace using prompts. Here’s what you’re going to ask your agent to do:
  1. Create a new Coda doc.
  2. Add a table.
  3. Populate it with sample tasks.
Now let’s give it a try.
Copy this prompt
Create a Coda Doc to track our team’s work. Add a projects section and a tasks table where tasks can be organized under projects and assigned to people. Include example data and share the link when you're done.
When the link appears, open it and explore your new doc. Not bad for one prompt.

Step 3. Ask a question

 Now that your agent has some context, let’s ask it a question.
Copy this prompt
How many projects do we have? Add the total to the top of the doc.

Step 4. Make it beautiful

 It’s time to make the page feel finished. Small visual touches make a doc easier to read and easier to share. You can give detailed prompts or simply ask the agent to beautify your doc, and it will add Coda features like:
  • Callouts
  • Icons
  • Emojis
Copy this prompt
Beautify and organize my doc with a title and subtitle, icons, friendly callouts, and a cover image. If needed, update column types to match the data.
Watch your doc level up with styling updates. Functional is good. Functional and delightful is even better.

Step 5. Mark all tasks complete

 Let’s finish strong. Instead of updating each row one by one in your task table, update them all at once.
Copy this prompt
Mark all task statuses as Complete.
Everything updates in a single sweep. A few prompts ago, you didn’t even have a doc. Now your doc is structured and styled. That’s Coda MCP in action.

You did it!

 Now that you’ve built your first doc with Coda MCP, imagine applying this to your work. Your agent can:
  • Summarize long docs and extract key information.
  • Generate more detailed trackers, project hubs, or briefs from notes.
  • Update tables based on changing inputs.
  • Refactor messy pages into organized systems.

Lightbulb icon
Quick tip
You can hand your agent any Coda page by sharing its URL. From there, it can summarize the content, reorganize it, add new pages or subpages, or help you turn it into something new. This is only the beginning.

Now what?

When you’re ready to put the same idea to work on something real, try one of these:

Was this helpful?

YesNo