Expose Flows & Sidekicks as MCP Server Tools

More detailed:
 

Summary
Allow external AI agents (Claude Code, Cursor, GitHub Copilot, etc.) to trigger Flows and invoke Sidekicks programmatically through Miro's MCP Server — enabling true end-to-end agentic workflows from code to canvas.

Problem
Today, Miro's MCP Server supports reading board content, creating diagrams (flowcharts, UML, ERD), managing docs, and working with tables. These are powerful building blocks, but they stop short of what's possible.

Flows and Sidekicks — Miro's most advanced AI capabilities — are only accessible through the Miro UI. This creates a disconnect: I can use Claude Code to generate an architecture diagram on a Miro board via MCP, but I cannot then trigger a Flow to refine it, apply branding, or generate accompanying documentation. I have to switch to the Miro UI, manually run the Flow, and visually verify the result.

For consultants and teams who build client deliverables (process diagrams, architecture maps, sales templates), this breaks the automation chain. The MCP Server acts as a data plane (read/write items) but not as a control plane (orchestrate intelligent workflows).

Proposed Solution
Add the following MCP tools to the Miro MCP Server:

1. flow_list — List available Flows for a board/team
2. flow_trigger — Trigger a Flow with specified input context (e.g., board items, text, URLs)
3. flow_status — Check the status/progress of a running Flow
4. flow_get_output — Retrieve the output/deliverables generated by a completed Flow
5. sidekick_invoke — Send a prompt to a Sidekick with board context and receive its response
6. board_screenshot / board_visual_check — Capture a visual snapshot of a board area so an external agent can verify the output

Use Cases
- Consultant workflow: Claude Code analyzes a codebase → creates an architecture diagram via MCP → triggers a "Brand & Polish" Flow that applies company styling → the agent verifies the visual result and iterates if needed
- Sales enablement: An AI agent creates a sales call template on a board → triggers a Flow that populates it with client-specific data → generates a presentation-ready deliverable
- Process documentation: An agent reads requirements from a doc → creates a process flow diagram → invokes a Sidekick to review it for completeness → triggers a Flow to generate accompanying documentation

Why This Matters
MCP was designed as a universal protocol for AI agents to interact with tools. Miro has embraced MCP — which is great — but currently only exposes the basic CRUD operations. Flows and Sidekicks are what make Miro's AI unique, and making them accessible via MCP would:
- Position Miro as the most complete AI-native visual workspace in the MCP ecosystem
- Justify the Business + AI Workflows plan for developer-heavy teams
- Enable workflows that no competitor currently offers
- Align with the MCP vision of interoperable, composable AI agents

I'm a paying Business plan customer upgrading to AI Workflows specifically because of these capabilities. Being able to orchestrate them from my existing AI coding tools would be a game-changer.