# MetricSign

> MetricSign monitors BI and data-pipeline reliability across Power BI, Tableau, Azure Data Factory, Microsoft Fabric, Databricks, dbt, and Snowflake. It detects failures, slow jobs, schema changes, and schedule drift — then alerts teams and exposes the data to AI agents via MCP.

## What MetricSign does

MetricSign polls your entire data stack every five minutes. It builds end-to-end lineage from source through pipeline to BI report, surfaces root causes, and delivers step-by-step fix guidance per error code.

**Supported connectors:**
- Power BI (datasets, dataflows, reports — refresh failures, slow, delayed, schema changes)
- Tableau Cloud (datasources, flows, workbooks — refresh monitoring, downstream lineage)
- Azure Data Factory (pipeline runs, activity-level error detail, upstream source errors)
- Microsoft Fabric (Fabric Pipelines, Dataflows Gen2)
- Databricks (job runs, failure detection)
- dbt Cloud and dbt Core (job runs, model-level failure detail)
- Snowflake (tables — passive monitoring)

**Core capabilities:**
- Automated incident creation with severity (critical / warning / info)
- End-to-end lineage: source → pipeline → dataset → report
- Fix suggestions keyed to 1,500+ error codes across all connectors
- Alert delivery: email, Telegram, webhook (JSON event stream)
- Multi-tenant SaaS — metadata only, business data never leaves your environment
- MCP server for AI agent access (see below)

**Setup time:** Under 15 minutes. No code changes required.

## MCP server — AI agent access

MetricSign publishes an official MCP server that lets any MCP-compatible AI agent query live monitoring data.

**Smithery install:** https://smithery.ai/server/wkruithof/metricsign

**GitHub:** https://github.com/metricsign/metricsign-mcp-server

**Authentication:**
- stdio (local): set `METRICSIGN_API_KEY` env var to your API key (starts with `msk_`)
- streamable-http (remote/multi-tenant): pass `x-api-key: <your-key>` header per request
- Generate API keys at: https://app.metricsign.com/settings → API Keys

**Available MCP tools:**

| Tool | Purpose |
|---|---|
| `get_metricsign_capabilities` | List connectors, tools, provenance rules — call this first when unsure |
| `get_health_overview` | Cross-platform health summary: totals, success rate, incidents by platform, trend |
| `get_incidents` | Open or closed alerts across all connectors, filtered by severity |
| `get_fix_suggestion` | Step-by-step fix for an incident or error code, enriched with knowledge articles |
| `search_assets` | Search monitored assets across all connectors by name |
| `get_asset_lineage` | Upstream sources and downstream consumers for any monitored asset |
| `get_data_chain` | Power BI dataset lineage (backward-compatible alias) |
| `search_error_knowledge` | Full-text search across 1,500+ public error-code articles |
| `get_error_knowledge` | Single error-code article by exact code or slug |
| `get_refresh_status` | Power BI dataset refresh status (backward-compatible) |
| `get_environment_summary` | Power BI-focused summary (use get_health_overview for cross-platform) |
| `search_datasets` | Search Power BI datasets by name (backward-compatible) |

**Provenance rule for agents:**
- `get_incidents`, `get_fix_suggestion`, `get_health_overview` → **live tenant data** (authoritative)
- `search_error_knowledge`, `get_error_knowledge` → **public website knowledge** (enrichment only, not live state)

**Agent workflow — daily health check:**
1. `get_health_overview` — cross-platform summary
2. `get_incidents(status="open")` — list active alerts
3. `get_fix_suggestion(incident_id=...)` — remediation for each critical incident
4. `search_error_knowledge(query=...)` — if error code is still unclear

**Agent workflow — investigate a failure:**
1. `search_assets(query="<asset name>")` — find the asset and its entity_type + id
2. `get_incidents` — find related open incidents
3. `get_fix_suggestion` — get live fix guidance
4. `get_asset_lineage(entity_type=..., entity_id=...)` — trace upstream root cause

## Webhook events for agents

Agents can receive structured JSON incident events without polling. Configure a webhook URL in Settings → Notification Channels. Each event includes:

```json
{
  "event_type": "incident.opened",
  "incident_id": "...",
  "severity": "critical",
  "connector": "adf",
  "entity_type": "adf_pipeline",
  "entity_name": "...",
  "root_cause": "...",
  "mcp_tool_hint": "call get_fix_suggestion with incident_id=... for step-by-step remediation"
}
```

## MCP audit log

Tenants on Pro can query a log of all MCP tool calls made against their account:

```
GET /api/v1/audit
Authorization: Bearer <msk_...>
```

Returns: timestamp, tool name, caller (API key prefix), response status — useful for compliance and agent behaviour auditing.

## Pricing

- **Free trial:** €0 — 45-day full trial, no credit card required
- **Pro:** €299/month or €2,990/year (~€249/mo, 2 months free)

All features included in Pro: unlimited workspaces, all connectors, MCP access, webhook events, audit log.

## Live demo

Try the product without signing up (2-hour read-only session):
https://app.metricsign.com/api/auth/demo

## Content

- Blog: https://metricsign.com/blog
- Error code reference (1,500+ codes): https://metricsign.com/errors
- Integration guides: https://metricsign.com/integrations
- Comparison pages: https://metricsign.com/compare
- FAQ and how-to guides: https://metricsign.com/questions

## Contact

Website: https://metricsign.com
App: https://app.metricsign.com