Product
Fanal sits between your pipeline monitoring and your stakeholders. Six specialised agents handle detection, lineage tracing, impact estimation, runbook generation, communication, and routing — fully autonomous, under 60 seconds.
How it works
Receives the raw webhook payload from BigQuery, dbt Cloud, or Airflow. Classifies the failure type (schema_drift, data_quality, dag_timeout, dependency_failure, job_failure) and scores blast radius 0.0–1.0 using downstream BFS traversal.
Tools
BigQuery MCP, Gemini 2.5 Flash
Outputs
failure_type, severity, blast_radius_score
Walks BigQuery column-level lineage using the MCP server. Performs BFS up to 5 hops downstream from the failed resource, identifying every dependent table, view, and materialisation.
Tools
BigQuery MCP, Gemini 2.5 Flash
Outputs
affected_tables, affected_views, hop_depth_map
Cross-references the lineage map against Looker dashboards, Vertex AI feature stores, and SLA-bound reports. Estimates financial exposure in dollars by multiplying affected revenue metrics by the failure window.
Tools
Looker API, Vertex AI API, Financial models
Outputs
affected_dashboards, affected_ml_models, financial_impact_usd
Generates a structured, failure-type-specific runbook covering triage, investigation, and remediation phases. Each step is assigned an owner audience (data engineer, DBA, platform team) and includes the exact commands or queries to run.
Tools
Deterministic templates, failure-type classifier
Outputs
runbook_steps[] with category, title, description, command
Writes five distinct, audience-specific messages — one each for data engineers, BI analysts, finance teams, ML teams, and executives. No technical jargon for non-technical audiences. Full stack context for engineers.
Tools
Gemini 2.5 Flash, Message templates
Outputs
5× stakeholder_message objects with audience, tone, channel
Routes each message to the correct Slack channel by audience role. Creates a Jira ticket with full incident context. Sends executive email for CRITICAL severity. Tracks acknowledgements and auto-escalates unacknowledged CRITICAL incidents after 15 minutes.
Tools
Slack API, Jira REST API, SendGrid
Outputs
delivery_receipts, jira_ticket_key, escalation_log
Integrations
Unique capabilities
Most observability tools stop at detection. Fanal goes further: the Impact agent calculates real revenue exposure by multiplying affected metric volume by time-to-resolution estimates. Finance receives a dollar figure, not a table name. This transforms Fanal from an engineering notification tool into a business risk management platform.
The Detector agent continuously monitors pipeline run times against a rolling 30-day baseline. When a pipeline is running 40%+ slower than its historical p95, Fanal fires a pre-incident warning — with predicted blast radius and affected stakeholders — before any failure occurs. You fix it before your stakeholders notice.
After an incident is resolved, Fanal automatically generates a structured postmortem: timeline, root cause, blast radius, MTTR, financial impact, and a prevention recommendation specific to the failure type. Delivered as Markdown to Confluence or Notion. One click to share.
Every Slack message includes an Acknowledge button. Fanal tracks whether each notified stakeholder has acknowledged. For CRITICAL incidents, if no acknowledgement is received within 15 minutes, the Coordinator agent automatically escalates to the next-level contact — up to the CFO or CTO — with full incident context.
Built with
| Layer | Technology |
|---|---|
| Orchestration | Google Agent Development Kit (ADK) |
| LLM | Gemini 2.5 Pro / Flash |
| Data lineage | BigQuery MCP Server |
| Backend | Django 5 + Django REST Framework |
| Frontend | Tailwind CSS + HTMX |
| Infrastructure | Google Cloud Run |
| Database | PostgreSQL (Cloud SQL) |
| Messaging | Slack Block Kit API |
| Ticketing | Jira REST API v3 |
14-day free trial · 100 incidents · No credit card required