Domain-agnostic by design. Works with any structured database.
Features
No SQL Required
Ask complex analytical questions in plain English. Provenance understands your intent and translates it into precise data operations automatically.
Conversational Data Analysis
Ask multi-step queries like "Compare energy consumption across our top 5 stores last quarter" and get instant answers.
Follow-Up Suggestions
Smart suggestion chips after every answer help you drill deeper without starting over.
Session Continuity
Full conversation history with the ability to edit and regenerate any previous query.
No Blind Trust
Every answer comes with a full provenance chain: what data was fetched, what transformations were applied, what SQL ran, and why. Analytics you can verify.
Full Execution Transparency
The Provenance button reveals the complete execution pipeline behind every answer.
Reasoning Transparency
Collapsible step-by-step AI thinking process so you can follow every decision.
Artifact Lineage
Every transformation is tracked in a Neo4j provenance graph — from source data to final report.
No Rigid Dashboards
The system dynamically generates the right visualization based on what the data calls for. No setup, no configuration, no training required.
Smart Visualization
Auto-decides the best chart type for each query — bar, line, pie, or something else entirely.
Interactive Data Tables
Sortable, searchable, paginated tables with automatic type detection for every result.
Real-Time Streaming
Progressive delivery via SSE — see results as they are computed, no loading spinners.
How It Works
A deterministic 6-step pipeline from question to verified answer.
Understand & Plan
Phase 1
WHAT HAPPENS
- Intent decomposer breaks question into atomic analytical motifs
- Property-based capability graph finds optimal path
- 44 atomic data operations available
- Contract validator checks plan against database schema
- Auto-replans up to 3 times if validation fails
Validate & Execute
Phase 2
WHAT HAPPENS
- Two-phase anti-poisoning prevents hallucinated data reuse
- Tasks run in parallel for maximum speed
- Full artifact lineage tracked in Neo4j provenance graph
- SAGA pattern orchestration with cooperative cancellation
- Heartbeat liveness and distributed tracing
Report & Stream
Phase 3
WHAT HAPPENS
- 3-stage LLM pipeline produces narrative + table + chart
- Separates expensive reasoning from cheap execution
- Real-time SSE streaming with conversational middleware
- Progressive delivery — no loading spinners
- Smart follow-up suggestions after every answer
What Our Users Say
Trusted by teams across energy, retail, and finance.
Maria K.
Energy Operations Director
"Provenance replaced three separate dashboards for us. Our team now gets consumption comparisons and efficiency benchmarks in seconds — and we can actually verify how every number was calculated."
Thomas R.
Head of Retail Analytics
"The provenance chain is a game-changer. When our CEO questions a number in a report, we can trace it back to the exact query, transformation, and source table. No more "trust me, the dashboard says so.""
Sofia L.
Financial Controller
"Our finance team went from writing SQL tickets for the data team to getting answers themselves in plain English. Revenue trends, cost ratios, segment analysis — all with full transparency into how Provenance arrived at each answer."
44
Atomic data capabilities orchestrated through a property-based capability graph.
6-Step
Deterministic pipeline: Understand, Plan, Validate, Execute, Report, Stream.
Real-Time
SSE streaming with conversational middleware — results delivered as they are computed.