How to use the Power Automate via FlowStudio MCP agent skill

What the skill makes the agent do

An agent can read and change Power Automate cloud flows only when it can reach a FlowStudio MCP server. This skill provides the plumbing for that connection: how to authenticate with an API key, how to discover which tools are available, and how to handle large responses.

It also includes a reusable helper script pattern in Python or Node.js, and it points to four more specialised skills for building, debugging, monitoring and governing flows. Those specialised skills depend on this one.

• Authenticate to the FlowStudio MCP endpoint with an API key or JWT token in the x-api-key header
• Discover tools in small bundles such as build-flow, debug-flow and monitor-flow instead of loading every schema at once
• Handle oversized responses cleanly
• Use a helper function to send MCP tool calls from Python or Node.js

Install the skill (ZIP download or Claude apps)

Download the skill ZIP from the page where you found this guide. Unzip it into one of these locations: ~/.claude/skills/ for your personal skills, or .claude/skills/ inside a project folder if you want the skill limited to that project.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/
• Per project: unzip into .claude/skills/ in the project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Get the requirements in place before you start

The skill cannot connect without a FlowStudio MCP subscription or a compatible Power Automate MCP server. You also need an API key or JWT token, which is sent in the x-api-key header and not as a Bearer token. Have your Power Platform environment name ready, for example Default-<tenant-guid>.

The MCP endpoint is https://mcp.flowstudio.app/mcp. In ChatGPT or claude.ai there is no key: you add https://mcp.flowstudio.app/mcp/oauth as a connector and sign in with Microsoft. The helper examples need Python 3 (standard library urllib) or Node.js 18 or newer.

• FlowStudio MCP subscription or compatible server
• API key or JWT token in the x-api-key header
• Power Platform environment name, e.g. Default-<tenant-guid>
• Python 3 (urllib) or Node.js 18+ for helper examples

Trigger the skill with natural requests

You do not call the skill by name. Describe the outcome you want and the agent should load it when it is relevant. For a first connection, ask the agent to set up the FlowStudio MCP connection or to list what it can do. The skill tells the agent to start with list_skills, which returns the available bundles, and then load the schemas for the bundle it needs with tool_search.

When a task is more specific, the matching specialised skill takes over. Requests to make or change a flow point to the build skill, requests to diagnose a failing run point to the debug skill, and tenant-wide health or tagging point to the monitoring and governance skills, which are for Pro+ subscribers.

• "Connect to FlowStudio and show me which tool bundles are available."
• "Load the tools I need to debug a failing run of my approval flow."
• "Find the tool for cancelling a run."
• "Write a small Python helper that calls the FlowStudio MCP endpoint."

Example: a first connection with the helper

The core helper sends a JSON-RPC tools/call request to the MCP endpoint with your token in the x-api-key header. The cold-start pattern is to call list_skills first, then tool_search with a skill query to load the schemas for one bundle. Replace the token placeholder with your own key and keep it out of shared code.

Safety notes: scripts and network access

This skill runs scripts and uses the network. The helper sends your token and flow data to the FlowStudio MCP endpoint. Use only the endpoint and credentials you are authorised to use.

The skill enables direct edits to running automations. Do not let an agent change live production flows without review. Check the connection references and the definition an agent proposes before you approve a change, and prefer a test environment for first attempts. Tool schemas may lag the server, so confirm the current shape with tool_search before calling a tool you have not used recently.

• Keep API keys out of shared code and chat logs
• Review every proposed change to a live flow before it runs
• Confirm tool parameters with tool_search when unsure
• Avoid PowerShell or shell scripts for building or updating flows, since nested JSON is easy to break

Tips for getting reliable results

Let the agent use list_skills and tool_search rather than guessing tool names. If the documentation and a real API response disagree, the API response is correct. Use the Python or Node.js helper for flow operations because JSON handling is safer than shell quoting for complex definitions.

• Start with a read-only request such as listing flows before asking for changes
• Keep the environment name and token handy so the agent can connect without guessing
• If a request is ambiguous, ask the agent to search tools by keyword, such as "cancel run"