What miroapp/miro-ai does
The miroapp/miro-ai MCP server gives AI coding assistants access to Miro boards. It supports workflows that move information between software projects and visual collaboration spaces, including reading board content, creating diagrams, generating code from designs, and synchronizing tasks with Miro tables.
The repository includes the MCP server plus client-specific packaging for several AI tools. Claude Code receives a plugin with bundled skills, Gemini CLI uses an extension, Codex uses a local plugin, and Kiro provides a power for design-to-code workflows. Other clients can connect through MCP configuration.
Supported skills include visualizing a codebase on a board, performing visual code reviews from pull requests or branch changes, and extracting Miro specifications into .miro/specs/. Content authoring such as diagrams, documents, tables, and stickies is handled by the MCP tools without requiring a separate skill.
How it works
The miroapp/miro-ai MCP server is available at https://mcp.miro.com/. Clients that support remote MCP servers can add that URL to their MCP configuration. Claude Code, Gemini CLI, Codex, and other supported integrations provide their own installation paths and may manage the MCP connection for you.
Authentication uses Miro OAuth. On first use, the client opens a browser, where the user signs in, grants access to requested boards, and returns to the AI tool. Access therefore follows the permissions associated with the connected Miro account.
Requests can include a Miro board URL. For example, an agent can summarize a board, create a login-flow diagram, or explain an architecture diagram using the board as context. The available capabilities cover both reading existing material and writing new content to boards.
Setup and configuration
For Claude Code, install the marketplace and plugin with /plugin marketplace add miroapp/miro-ai followed by /plugin install miro@miro-ai. Gemini CLI can install the repository as an extension with gemini extensions install https://github.com/miroapp/miro-ai. Kiro installation uses its Powers panel and the powers/code-gen power from GitHub.
Agent Skills can be installed interactively with npx skills add miroapp/miro-ai, listed with --list, or selected with --skill. For clients such as Cursor, VS Code with Copilot, and Windsurf, add the hosted MCP URL to the client configuration.
Restart the relevant client when the installation instructions require it, then authenticate when prompted. Use only one Miro connection method in a client. A plugin or extension combined with a manually configured server can create duplicate tools and separate OAuth sessions.
Enterprise Miro accounts require an administrator to enable the MCP server before users can connect.
Tools and capabilities
The miroapp/miro-ai MCP server supports:
- Reading board content to supply context to an AI workflow.
- Creating flowcharts, sequence diagrams, and entity-relationship diagrams from descriptions or code.
- Turning board designs and wireframes into working code.
- Moving task information between AI conversations and Miro tables.
- Extracting documents, diagrams, prototypes, tables, and images into project specifications.
- Producing visual explanations of codebases and architecture.
- Reviewing code visually from pull requests, local changes, or branch comparisons.
Limitations and notes
The repository describes the MCP server as beta software. Availability and behavior may depend on the connected client, the user's Miro permissions, and board access granted during OAuth.
Miro API access is permission-based and rate limited. Enterprise deployments also depend on administrator approval.
Client-specific plugins and extensions include bundled skills, while a manually configured MCP connection provides the server capabilities without necessarily installing those skills. Check the integration method for the client you plan to use.