Flowstudio Power Automate MCP

Power Automate via FlowStudio MCP foundation

Connect an agent to the FlowStudio Power Automate MCP server, verify authentication, discover current tool schemas, choose the right workflow skill, and handle large MCP responses without flooding context.

When to invoke

• "Connect GitHub Copilot to FlowStudio MCP for Power Automate."
• "Set up the MCP helper for Power Automate flows."
• "List the FlowStudio MCP tools and skill bundles."
• "My FlowStudio MCP auth is failing with 401 or 403."
• "How should I handle a huge get_live_flow or connector response?"

Prerequisites and context

• Requires a FlowStudio MCP subscription or compatible Power Automate MCP server: https://mcp.flowstudio.app.
• MCP endpoint: https://mcp.flowstudio.app/mcp.
• API key / JWT token: send the plain token in the x-api-key header, not Authorization: Bearer.
• Token placeholder in examples: <YOUR_JWT_TOKEN>; keep _TOKEN semantics intact when adapting scripts.
• Power Platform environment name, usually Default-<tenant-guid>, found through list_live_environments or a list_live_flows response.
• Use a timeout of at least 120 seconds for large calls such as get_live_flow_run_action_outputs.

Workflow skill routing

Pick by user intent, not by individual tool names. The same underlying MCP tools appear in multiple workflows.

flowstudio-power-automate-build and flowstudio-power-automate-debug can both call update_live_flow, get_live_flow, and run-error tools; build works forward from desired behavior, while debug works backward from a failed run. flowstudio-power-automate-monitoring and flowstudio-power-automate-governance both call Store tools; monitoring reads health, while governance writes metadata.

Source of truth

If docs disagree with a live response, the API wins. Before invoking a tool you have not used recently, call tool_search to confirm the current schema.

Tool discovery

FlowStudio MCP v1.1.5+ exposes non-billable meta-tools that avoid loading all 30+ schemas.

Current common bundles:

skills = mcp("list_skills", {})
debug_tools = mcp("tool_search", {"query": "skill:debug-flow"})
selected = mcp("tool_search", {"query": "select:get_live_flow,update_live_flow"})

MCP helper patterns

Use Python with urllib.request when you want a dependency-free helper. Use Node.js 18+ when the repository is JavaScript or TypeScript and native fetch, JSON.stringify, and JSON.parse fit the stack. Avoid PowerShell for flow operations because ConvertTo-Json -Depth silently truncates nested definitions. Use cURL or Bash only for tiny smoke tests; nested JSON escaping is fragile.

Python helper

import json, urllib.request

TOKEN = "<YOUR_JWT_TOKEN>"
MCP = "https://mcp.flowstudio.app/mcp"

def mcp(tool, args, cid=1):
    payload = {"jsonrpc": "2.0", "method": "tools/call", "id": cid,
               "params": {"name": tool, "arguments": args}}
    req = urllib.request.Request(MCP, data=json.dumps(payload).encode(),
        headers={"x-api-key": TOKEN, "Content-Type": "application/json",
                 "User-Agent": "FlowStudio-MCP/1.0"})
    try:
        resp = urllib.request.urlopen(req, timeout=120)
    except urllib.error.HTTPError as e:
        body = e.read().decode("utf-8", errors="replace")
        raise RuntimeError(f"MCP HTTP {e.code}: {body[:200]}") from e
    raw = json.loads(resp.read())
    if "error" in raw:
        raise RuntimeError(f"MCP error: {json.dumps(raw['error'])}")
    text = raw["result"]["content"][0]["text"]
    return json.loads(text)

Node.js helper

const TOKEN = "<YOUR_JWT_TOKEN>";
const MCP = "https://mcp.flowstudio.app/mcp";

async function mcp(tool, args, cid = 1) {
  const payload = {
    jsonrpc: "2.0",
    method: "tools/call",
    id: cid,
    params: { name: tool, arguments: args },
  };
  const res = await fetch(MCP, {
    method: "POST",
    headers: {
      "x-api-key": TOKEN,
      "Content-Type": "application/json",
      "User-Agent": "FlowStudio-MCP/1.0",
    },
    body: JSON.stringify(payload),
  });
  if (!res.ok) {
    const body = await res.text();
    throw new Error(`MCP HTTP ${res.status}: ${body.slice(0, 200)}`);
  }
  const raw = await res.json();
  if (raw.error) throw new Error(`MCP error: ${JSON.stringify(raw.error)}`);
  return JSON.parse(raw.result.content[0].text);
}

For older Node.js, replace fetch with https.request from the standard library or install node-fetch.

Connection verification and errors

skills = mcp("list_skills", {})
print(f"Connected — {len(skills)} skill bundles available:",
      [s["name"] for s in skills])

Expected output:

Connected — 6 skill bundles available: ['build-flow', 'create-flow', 'debug-flow', 'monitor-flow', 'discover', 'governance']

Oversized responses

Some responses can overflow the agent context window.

When a harness spills output to a file such as tool-results/mcp-flowstudio-describe_live_connector-NNNN.txt, parse the double-wrapped payload before filtering:

import json
with open(path) as f:
    raw = json.loads(f.read())
payload = json.loads(raw[0]["text"])

$payload = ((Get-Content $path -Raw | ConvertFrom-Json)[0].text) | ConvertFrom-Json

Rules of thumb:

1. Extract one operationId, one action's outputs, or the specific fields needed; do not echo whole connector Swagger documents.
2. Always pass actionName to get_live_flow_run_action_outputs; add iterationIndex for foreach actions when you need one repetition.
3. Reuse a spill file within a session instead of refetching the same connector swagger.
4. Do not grep spill files directly for JSON keys such as "OperationId":; strings are JSON-escaped as \"OperationId\":. Parse first, then filter.
5. Summarize name + state + trigger for flow lists and actionName + status + code for run errors unless the user asks for raw JSON.

conn = mcp("describe_live_connector", {"environmentName": ENV, "connectorName": "shared_sharepointonline"})
op = conn["properties"]["swagger"]["paths"]["/datasets/{dataset}/tables/{table}/items"]["get"]
print(op["operationId"], "—", op.get("summary"))

Progressive disclosure and bundled resources

Read bundled references only when the current task needs them.

• references/MCP-BOOTSTRAP.md: endpoint, auth, JSON-RPC request/response format; read first for setup.
• references/tool-reference.md: response shapes and behavioral notes; use tool_search for current parameters.
• references/action-types.md: Power Automate action type patterns.
• references/connection-references.md: connector reference guide.

FlowStudio vocabulary and aliases

Keep these operational phrases recognizable when discussing FlowStudio MCP: oversized-response, use-case, most-likely-needed, request-response, built-in, async/await, JavaScript/TypeScript, pip install, smoke-test, error-prone, top-level, inputs/outputs, Starting/stopping, tools/list, query: "<keywords>", "cancel run", select:<toolname>, 401/403, x-api-key: <JWT>, get_live_flow_runs, and get_live_flow_run_error.

Output template

## FlowStudio MCP foundation result

**Status:** connected | needs auth | blocked
**Endpoint:** `https://mcp.flowstudio.app/mcp`
**Environment:** `<Default-tenant-guid or unknown>`

### Discovery
- `list_skills`: <bundle count> bundles found
- `tool_search`: `<query>` → <tool count> schemas loaded

### Helper
- Language: Python | Node.js
- Auth header: `x-api-key`
- Timeout: `<seconds>`

### Next workflow
- Skill: `<flowstudio-power-automate-build|flowstudio-power-automate-debug|flowstudio-power-automate-monitoring|flowstudio-power-automate-governance>`
- Reason: <user intent matched>

Quality gate

• The MCP endpoint is https://mcp.flowstudio.app/mcp.
• The token is sent as x-api-key, not Authorization: Bearer.
• list_skills or tool_search was used before relying on a workflow tool schema.
• The helper uses JSON-RPC tools/call with params.name and params.arguments.
• Large outputs are parsed, filtered, and summarized instead of echoed wholesale.
• Workflow routing is based on user intent and notes Pro+ requirements for monitoring or governance.

References

• FlowStudio MCP: https://mcp.flowstudio.app
• FlowStudio MCP endpoint: https://mcp.flowstudio.app/mcp
• FlowStudio MCP subscription page: https://mcp.flowstudio.app.
• Expression error in child flow: https://github.com/ninihen1/power-automate-mcp-skills/blob/main/examples/fix-expression-error.md
• Data entry, not a flow bug: https://github.com/ninihen1/power-automate-mcp-skills/blob/main/examples/data-not-flow.md
• Null value crashes child flow: https://github.com/ninihen1/power-automate-mcp-skills/blob/main/examples/null-child-flow.md