Domain Scanner
Inventories workspaces, lakehouses, warehouses, semantic models, and owners across your tenant.
- Live Fabric API
- Lineage hints
- Quality signals
RECON · v1.0 · LIVE
Every signal.
Every domain.
One trusted map.
FabricRecon scans your Microsoft Fabric estate as living terrain. Probes sweep each domain, anomalies rise as signal pillars, and agent pools reconcile drift — so your team acts on a single, explainable topography of the data they own.
- 32
- probes shipped
- 5
- specialist agents
- 4
- data domains
- 1
- unified data map
The intelligence stack
Reconnaissance, reconciliation, and synthesis — built natively for Fabric.
Four cooperating capabilities turn a sprawling data estate into a single, trusted topography. Each one is observable, scriptable, and lives entirely inside your tenant.
Probe Runner
Dispatches targeted profiling, freshness, duplicate, and overlap probes against each domain.
- 32 default probes
- Custom playbooks
- Retry + replay
Agent Pool
Specialist AI agents compare records, score matches, and log evidence for every reconciliation call.
- GPT-4o reasoning
- Confidence scores
- Audit trail
Synthesis
Composes probe output + agent decisions into a unified, explainable data map with next-best actions.
- Navigable map
- Open findings queue
- Briefing digest
The daily briefing
A morning standup written by the agents that worked overnight.
FabricRecon publishes a ranked briefing every morning — composite scores, narrative explanations, and the exact probes and agents that produced each finding. Open the top item, agree with the recommendation, and move on.
- Composite score across freshness, drift, duplicates, and impact
- Narrative built from probe evidence — not a wall of metrics
- One-click Acknowledge, Investigate, or Dismiss.
Inside the workload
Three native items. One operating model.
FabricRecon ships as Fabric workload items you can govern, schedule, and audit alongside every other notebook, pipeline, and warehouse in your tenant.
01
Recon Job
Scan the estate. Surface the signal.
Runs domain reconnaissance jobs that inventory Fabric assets, detect schema and lineage hints, profile quality signals, and queue follow-up probes for suspicious gaps or overlap.
4 domains · 32 probes
02
Setup Job
Configure reconnaissance once. Operate continuously.
Connect live API endpoints, define domains, register probe playbooks, set confidence thresholds, and configure agent pools for your tenant-native Fabric deployment.
5-step wizard · 3 minute install
03
Synthesis
Turn findings into a unified map.
Combines scan output, probe evidence, and agent decisions into an explainable data map that shows trusted matches, unresolved conflicts, and next-best actions.
Composite scores · evidence trails
2:30 product tour
Watch FabricRecon discover, probe, and reconcile — in real time.
A narrated walk-through of the daily briefing, finding detail, probe bank, agent pool, run history, and setup wizard. Every screen is the actual product running against rich demo data.
The recon loop
From estate scan to trusted map — in a continuous loop.
FabricRecon doesn't run once. Every probe outcome feeds the next scan, every reconciliation tunes the next agent, and every briefing builds on the last.
- 01
Scan
Domain Scanner sweeps workspaces, lakehouses, warehouses, and semantic models. Schema, lineage, ownership, and freshness signals flow into the recon graph.
- 02
Probe
The Probe Runner dispatches focused tests — duplicates, overlaps, keys, drift, semantic similarity — against every domain that warrants a closer look.
- 03
Reconcile
Specialist AI agents inspect probe output, score candidate matches, resolve conflicts at confidence, and log evidence for everything they touch.
- 04
Synthesise
Findings collapse into a single navigable data map with composite scores, narrative explanations, and recommended next probes.
Inside the app
Every screen, designed for the operator on shift.
Built for the data engineers, analytics leads, and platform owners who have to answer for the data they ship.
Built on Microsoft Fabric
Tenant-native. Agent-powered. Operator-grade.
Runs in your tenant
FabricRecon is a sideloaded Fabric workload — capacity, identity, networking, and governance are yours. No data leaves your boundary.
AI Foundry · GPT-4o
Specialist agents reason against probe evidence using Azure AI Foundry. Every decision is logged with prompts and reasoning chain.
Native item model
Recon Job, Setup Job, and Synthesis are first-class Fabric items — schedule them, share them, govern them with the same model as every other asset.
Live API endpoints
Probes connect to real Fabric REST and Onelake endpoints. Mock mode is available, but production points at your actual capacity.
FAQ
Questions we get every week.
What is FabricRecon?
FabricRecon is a native Microsoft Fabric workload for AI-powered reconnaissance and reconciliation across data estates. It discovers domains, runs probes, coordinates agent pools, and synthesises findings into a unified data map.
Does FabricRecon move my data out of Fabric?
No. FabricRecon is a sideloaded workload that operates against your Fabric tenant and live API endpoints. Identity, capacity, and governance stay yours.
How do agent pools actually work?
Each pool dispatches a specialist agent (drift, duplicate, freshness, lineage, reconcile) against probe evidence. Agents score candidate matches, resolve at high confidence, and queue everything else for human review with a full evidence trail.
Can I define my own probes?
Yes. Probe playbooks let teams target freshness, schema drift, duplicate keys, semantic overlap, lineage clues, and any reconciliation check specific to their domains.
What does the briefing actually contain?
A composite-scored, ranked list of findings with a one-paragraph narrative for each: what changed, what it impacts, which probes detected it, which agents acted, and the recommended next action.
Who builds FabricRecon?
FabricRecon is built by Datachain Consulting Pty Ltd — a Sydney-based data engineering consultancy specialising in Microsoft Fabric and enterprise data platforms.
Ready when you are
Bring FabricRecon into your tenant.
The workload sideloads in minutes. The first briefing lands the morning after your first scan.