MCP Fleet Overview

Aggregate the last 30 days of MCP telemetry across every server and category into a single ranked table of calls, errors, connections, and active users.

Best for

Platform PMs · DevOps · Leadership

Runs with

BigQuery · Fleet ranking · Connection status

Core outputs

• Ranked server table
• Error rates
• Connection status
• Active users

Capability requirements

MCP

What it solves

No quick, repeatable snapshot exists. Someone has to go looking, table by table.

One-shot snapshot of the entire fleet. Not a table someone had to go find.

Dormant connections and error concentration, surfaced instantly.

A typed table, ready to pipe straight into a dashboard.

Workflow

1. Run one read-only BigQuery aggregation across the MCP catalog.
2. Rank servers by call volume with error and connection-status columns.
3. Return a typed, chart-ready records table for the fleet.

Implementation

Review the underlying FML blueprint — the sessions, tools, and typed schemas that define this workflow — and see exactly how it is instructed for repeatable execution.

FML Blueprint

mcpops-active.fml

require mcp BigQuery

session("mcp_analytics") {
    use mcp BigQuery

call("execute_sql_readonly") -> vars:bq_result {
        projectId = "barndoor-production"
        query     = "SELECT mcp_name, mcp_category, COUNT(DISTINCT tenant_id) AS tenant_count, SUM(total_tool_calls) AS total_calls, SUM(total_errors) AS total_errors, SAFE_DIVIDE(SUM(total_errors), NULLIF(SUM(total_tool_calls), 0)) AS error_rate, MAX(last_tool_call) AS last_tool_call, COUNTIF(connected_but_never_called) AS connections_never_called, SUM(active_users) AS total_active_users, COUNT(DISTINCT IF(status = 'Active', mcp_id, NULL)) AS active_connections, COUNT(DISTINCT IF(status = 'Pending', mcp_id, NULL)) AS pending_connections, COUNT(DISTINCT IF(status = 'Error', mcp_id, NULL)) AS error_connections FROM `barndoor-production.gtm_analytics.mcp_block` WHERE DATE(last_tool_call) >= DATE_SUB(CURRENT_DATE(), INTERVAL 30 DAY) OR last_tool_call IS NULL GROUP BY mcp_name, mcp_category ORDER BY total_calls DESC LIMIT 50"
    }

+ The BigQuery query has executed successfully. Here is the raw JSON result:
      {{ .vars.bq_result | json }}

- Parse the BigQuery result into the structured schema below.
      The data contains an array of rows where each row has columns in the exact order as the schema fields (mcp_name, mcp_category, tenant_count, total_calls, total_errors, error_rate, last_tool_call, connections_never_called, total_active_users, active_connections, pending_connections, error_connections).
      Convert string representations of numbers to their appropriate int/float types.

schema {
        records: {
            mcp_name: string
            mcp_category: string
            tenant_count: int
            total_calls: int
            total_errors: int
            error_rate: float
            last_tool_call: string
            connections_never_called: int
            total_active_users: int
            active_connections: int
            pending_connections: int
            error_connections: int
        }[]
    }
}