Analytics MCP Server

Governed analytics platform for AI agents — metrics, dashboards, funnels, cohorts, anomaly detection, forecasting, and policy enforcement. 28 tools with full audit trail.

Architecture

MCP Analytics Architecture

Tools (28)

Discovery (5)

| Tool | Purpose | |------|---------| | list_data_sources | List connected data sources (warehouse, product DB, event stream) | | list_datasets | List available datasets/tables with row counts | | describe_dataset | Schema, columns, types, PII flags, freshness | | list_metrics | List defined metrics with owners and certification | | get_metric_definition | Full metric: formula, aggregation, dimensions |

Querying (5)

| Tool | Purpose | |------|---------| | query_metric | Query metric over time range → time-series data | | breakdown_metric | Break metric by dimension (country, plan, device) | | compare_metric | Compare metric across two periods | | query_events | Query raw events with filters | | query_report | Run a saved/named report |

Analysis (6)

| Tool | Purpose | |------|---------| | analyze_funnel | Conversion rates through funnel steps | | analyze_cohort | Weekly/monthly cohort retention | | detect_anomalies | Find anomalies in recent metric data | | forecast_metric | Forecast metric N days forward with confidence | | explain_change | Dimension attribution for metric changes | | generate_insight_summary | AI-generated highlights + recommendations |

Dashboard Building (5)

| Tool | Purpose | |------|---------| | list_dashboards | List all dashboards | | get_dashboard | Get full dashboard with widgets | | summarize_dashboard | Key takeaways from a dashboard | | create_dashboard | Create a new dashboard | | add_widget | Add chart/number/table/funnel widget | | publish_dashboard | Publish dashboard to team |

Governance (5)

| Tool | Purpose | |------|---------| | validate_analytics_policy | Check if action is allowed by policy | | check_export_risk | Assess PII/row-count risk of export | | request_data_access | Request approval for restricted data | | get_query_audit_trail | View audit log of all queries | | get_segments / query_segment | List and query user segments |

Installation

cargo install mcp-analytics

Configuration

No configuration required — starts with seeded demo data:

3 data sources, 4 datasets, 8 metrics
5 user segments, 2 funnels, 1 pre-built dashboard
Policy enforcement active by default

Policy Boundaries (enforced)

| Policy | Default | |--------|---------| | raw_sql_allowed | false | | pii_column_access | denied | | row_limit | 10,000 | | customer_level_export | requires_approval | | employee_level_export | denied | | financial_metric_access | approved_only | | external_sharing_allowed | false | | metric_certification_required | true |

Client Configuration

Claude Desktop / Kiro / Cursor

{
  "mcpServers": {
    "analytics": {
      "command": "mcp-analytics",
      "args": []
    }
  }
}

End-to-End Example: Building a Revenue Dashboard

Agent: "Create a revenue dashboard for the exec team"

1. list_metrics → finds MRR, ARPU, Churn Rate
2. validate_analytics_policy(action="financial_access") → checks permission
3. create_dashboard(name="Revenue Dashboard")
4. query_metric(metric_id="m-mrr", days=30) → gets time-series
5. add_widget(dashboard_id, type="line_chart", title="MRR Trend", metric_id="m-mrr")
6. breakdown_metric(metric_id="m-mrr", dimension="plan") → by plan tier
7. add_widget(dashboard_id, type="bar_chart", title="MRR by Plan")
8. forecast_metric(metric_id="m-mrr", horizon_days=14)
9. add_widget(dashboard_id, type="line_chart", title="MRR Forecast")
10. detect_anomalies(id="m-churn") → finds spike
11. publish_dashboard(id=dashboard_id) → visible to team

Governance Model

The analytics server enforces policy at every step:

Pre-query validation — validate_analytics_policy checks before data access
Export risk assessment — check_export_risk flags PII and large exports
Approval workflow — request_data_access for restricted resources
Full audit trail — every query logged with timestamp, params, result
No raw SQL — prevents injection and uncontrolled data access

MCP Server Manifest

server_id = "mcp_analytics"
display_name = "Analytics"
version = "1.0.0"
domain = "business-systems"
risk_level = "medium"
writes_allowed = "gated"

License

Apache-2.0

Part of the ADK-Rust Enterprise MCP server ecosystem.

Built with ❤️ by Zavora AI

MCP Servers