Miro MCP Server — Frequently Asked Questions

What is the Miro MCP Server, and what can it do?

The Miro MCP (Model Context Protocol) Server is a secure gateway that lets MCP-compatible AI tools work with your Miro boards. In practice, an AI agent can:

• Read and summarize board content (sticky notes, shapes, frames, text, cards, docs, tables, comments).
• Create and update content — build structured layouts with frames, sticky notes, shapes, text, cards, docs and tables, and add images from public URLs.
• Generate diagrams from code or text, such as architecture, sequence, and entity-relationship (ERD) diagrams.
• Act on feedback — read and resolve comments and turn discussion into next steps on the canvas.

A helpful way to think about it: the MCP Server is the secure gateway, and MCP Clients are the specific integrations that connect individual AI tools (Claude, Cursor, GitHub Copilot, and others) to Miro.

Which AI tools work with the Miro MCP Server?

Any MCP-compatible client that supports remote connections and OAuth 2.1. Documented setups exist for Claude and Claude Code, Cursor, VS Code + GitHub Copilot, Microsoft 365 Copilot (via the Miro agent for Copilot and the Copilot Cowork Miro plugin), Windsurf, Gemini CLI, Lovable, Replit, Kiro, and Amazon Q, among others. If your tool has an MCP client but can't connect, confirm it supports remote MCP servers and that you're on the latest version.

How do I connect my AI tool to Miro?

The OAuth flow is the same across tools; only the UI differs. In general:

1. In your tool's connector or MCP settings, add Miro (or "Add MCP Server") and enter the server URL: https://mcp.miro.com.
2. Complete the Miro OAuth flow and select the team where your boards live.
3. Drop a full board URL into a prompt to start — for example, "Summarize the content on this board: <board-URL>."

For Claude specifically, use the + → Add connectors menu, search for Miro, connect, and complete OAuth. Step-by-step guides for each tool are in Miro's developer documentation.

Why can't I use Miro MCP within my Enterprise Org?

Contact your Admin. On Enterprise plans the MCP Server may be disabled by default so a company admin must turn it on. This means it won't work even if you request access via another tool such as Claude via Miro Connector.

Admins enable it under Admin settings → Apps & Integrations → MCP, then choose to enable it Org wide or for selected teams only:

• Organization — any user may authorize MCP for boards they can already access.
• Selected team(s) — only users in those teams may authorize MCP.

If you don't see the MCP section or the toggle, it typically means MCP hasn't been enabled for your organization or your specific team yet, or you don't have admin rights. Contact your Miro admin or account team.

Which plans include the MCP Server?

The MCP Server is available across Miro plans, with Enterprise requiring admin approval before users can install and use it (and disabled by default until an admin enables it).

Is Miro MCP Secure?

Access is governed by your existing Miro permissions. Authentication uses OAuth 2.1 with dynamic client registration, and the AI agent can only see and act on boards the authenticated user can already access. Enabling MCP respects your existing board-level access controls and enterprise security standards.

If a user loses access to a board, can a previously issued token still read it? Does MCP support revocation?

Access is evaluated against the user's current Miro permissions rather than granting a standing copy of your data.

Where is my data hosted, and does using MCP move it to another region?

Miro processes content securely following its enterprise standards and data-residency model. For a precise statement about your organization's residency region and how MCP traffic is handled relative to it, ask your account team for Miro's official data-residency and security documentation.

What admin controls and audit visibility are available?

Admins control availability at the organization level and choose which teams can use MCP; board-level permissions continue to apply on top of that. For questions about auditing which users are using MCP, restricting which tools or clients may connect, human-in-the-loop confirmation for actions, and rate/quota controls, the specifics are evolving — your account team can share the current governance capabilities and roadmap if needed for a security review.

Why do I have to re-authenticate, and can I connect more than one team at once?

Today the Miro MCP connection is 1:1 — each user is authenticated into one team at a time, even across multiple clients. If you authorize a different team (or a different client), the previous team's connection is superseded, and an agent trying to reach a board in the no-longer-active team will return an error such as "This board belongs to a different team than your current authorization."

The fix: re-authenticate while viewing a board in the team you want to use, which resets the connection to that environment. If you frequently work across teams, keep this in mind when planning workflows. Broader multi-team support is a frequently requested improvement.

My AI tool says it can't find my board. What's wrong?

Make sure the board URL is included in your prompt, that you're authenticated into the same team where the board lives, and that you already have access to that board in Miro. Including the full URL as a habit ("prompt hygiene") is a good idea.

What can and can't the MCP create today — sticky notes, connectors, images, custom shapes?

The MCP works with a curated set of Miro objects: frames, sticky notes, shapes, text, cards, docs, tables, and images from public URLs, plus generated diagrams. It is a purposeful abstraction rather than a full mirror of the Miro REST API, so some things behave differently or aren't available yet. Capabilities are expanding over time, so check the current tools list in Miro's developer documentation for what's supported now, and note that the REST API may cover use cases the MCP doesn't yet.

How do I get better, cleaner visual output when an AI builds a board?

A few practices consistently help:

• Be specific about layout and placement — tell the agent where to put content (for example, an empty area to the side) so it doesn't overlap existing work.
• Start scoped — one clear artifact or frame at a time tends to produce cleaner results than a large multi-board request in a single shot.
• Iterate — reading and summarizing existing boards is a strength; generating polished visual layouts from scratch may need a couple of refinement passes.
• For complex or highly formatted content, you can try to build your own custom skill after you've seen certain prompts work.