Get started
Fabric Desktop by KratosBI documentation
A new name. The same foundation. KratosForge is the upcoming name for Fabric Desktop by KratosBI. These guides preserve the app’s current labels and paths.
Fabric Desktop by KratosBI runs Microsoft Fabric data engineering work on your own PC (notebooks, Spark SQL, and a metadata-driven medallion framework) and writes the results straight into OneLake. Start with the guide for what you're doing.
Start here#
| I want to | Read |
|---|---|
| Install the app on my PC | Install |
| Back up my PC's notebooks and settings | Back up and restore |
| Roll it out to many PCs | Enterprise install |
| Set it up in Azure and Fabric for my organization | Configure Azure and Fabric, then Permissions |
| Learn what every page does | User guide |
| Write and run notebooks and SQL on my PC | Notebooks and SQL |
| Query a folder of data files with SQL on my PC | Local data projects |
| Compare a run on my PC with the same run in Fabric, or benchmark both | Side by side |
| Know where a CU or cost figure comes from and how far to trust it | Capacity and cost evidence |
| Share optimization results across every installation in my organization | Optimize |
| Build Bronze, Silver and Gold from config | Metadata framework guide |
| Set it up for several teams | Working across teams, The platform, Data mesh, Governance policies and the governance guide |
| Watch and manage every OneLake shortcut: health, access, changes, requests | Shortcut command center |
| Check a Fabric pipeline definition on my PC | Pipeline definitions |
| See which pipeline activities need Fabric capacity | Pipeline activity compatibility |
| Test a pipeline's control flow on my PC | Running pipelines on your PC |
| Fix an error | Troubleshooting |
| Review the app for security approval | Security evidence pack |
| Prove the app works with my tenant before a pilot | Live testing |
| Suggest a feature or vote on one | Submitting ideas |
All documents#
| Document | For | What it covers |
|---|---|---|
| install.md | Everyone | Requirements, download, first start, Windows blocking the engine, updates and channels, where files live, uninstalling |
| backup.md | Everyone | Backing up and restoring notebooks, settings, schedules and history on a PC |
| enterprise-install.md | Admins | Deploying with Intune or Configuration Manager, managed updates, rollback, failed updates, network access |
| azure-setup.md | Admins | Fabric tenant settings, sign-in policies, the optional policy file, workspaces and capacity, access, optional extras, a check that it works |
| permissions.md | Admins | How sign-in works with no app registration; what each feature uses; an optional own app registration and its permissions, in alphabetical order; what the agent needs |
| telemetry.md | Everyone, admins | Opt-in usage data: every field sent, what is never sent, where it goes, the policy keys |
| evidence-pack.md | Security reviewers, IT | Everything a security review needs in one place: data flows, documents, build and release evidence, and what is still open |
| security.md | Admins, security reviewers | The threat model: what is protected, trust boundaries, how updates are checked, and what is still open |
| hardening/README.md | Security reviewers, platform owners | The enterprise hardening audit: scorecard, findings, backlog and release gates |
| user-guide.md | Everyone | Every page: Home, Create, Workspaces, OneLake (catalog, shortcuts, access), Data mesh, Fabric, Monitor, Settings, the command bar and the bell |
| notebooks.md | Engineers, analysts | Notebooks on your PC, Fabric code that works here, publishing, Run in Fabric, schedules, SQL query |
| sql-compatibility.md | Engineers, analysts | Checking SQL against Fabric Warehouse, SQL analytics endpoint, SQL database and Spark before running it; the rules and their sources |
| sql-execution.md | Engineers, analysts, IT | How SQL page queries run on this PC: Stop, time and memory limits, paging, CSV/Parquet export, what a query may change, audit |
| local-data.md | Engineers, analysts | Local data projects (Preview): folders of files as SQL tables on your PC, naming rules, limits, where it is kept |
| side-by-side.md | Engineers, analysts | Running a notebook on this PC and in Fabric side by side, how results are checked, and the TPC-H benchmark |
| finops/capacity.md | Engineers, platform owners | The capacity each environment is on, whether it is running, and what to do when it can't be read |
| finops/risk-check.md | Engineers, FinOps, IT | What the risk check before a run looks at, its results, settings and acknowledgements |
| finops/usage-import.md | Engineers, capacity admins, FinOps | Import a Capacity Metrics app export as measured CU and compare it with this app's estimates; import an Azure Cost Management export for what each capacity was billed |
| optimize/README.md | Engineers, platform owners, admins | The central lakehouse (Preview): setting it up, permissions, what each installation writes, who can see what, the policy key |
| finops/cost-evidence.md | Engineers, platform owners | Optimize > Cost and savings figures: measured, estimated, scenario and unknown, what is counted, units, and the measurement schema |
| prove/how-proofs-work.md | Engineers, analysts | Proving a notebook: pinned inputs, sandboxed outputs, how results are verified, the scorecard and receipts (Preview) |
| metadata-guide.md | Engineers | The metadata framework step by step, from a new project to Prod, and a reference for every setting |
| platform.md | Platform owners | The central platform: one key system, one landing zone, the catalog; platform admins, requests, the Platform page; creating, joining, moving keys |
| governance.md | Platform owners, team leads | Governance policies (Preview): the policy registry, drafts and versions, enterprise to entity scopes, mandatory policies, governance.toml, who can change what |
| governance-guide.md | Platform owners, team leads | Governance guide (Preview): draft to result, worked examples (naming, classification, ownership), what can be enforced and by whom |
| data-mesh.md | Team leads, engineers | Domains, workspace designations, data products, links between teams, catalog text |
| shortcuts.md | Engineers, platform owners | The shortcut command center: health checks, who can read a shortcut, shortcut rules, change history with undo, repointing, requests |
| source-control.md | Engineers, analysts | Source control (Preview): clone, diffs, branches, commit, pull and push, the secret check, what it doesn't do yet |
| orchestration/pipeline-definitions.md | Engineers | Reading, checking and converting Fabric pipeline definitions on your PC (Preview) |
| orchestration/activity-registry.md | Engineers | Which pipeline activities run on your PC, are only checked, or need Fabric capacity (Preview) |
| orchestration/local-runs.md | Engineers | Running a pipeline's control flow on your PC with mocks; no capacity used (Preview) |
| teams.md | Team leads, platform owners | Roles, Git workflow, environments and workspaces, shared keys, feedback at scale |
| troubleshooting.md | Everyone | Each error message, what it means and what to do |
| release-management/README.md | Maintainers, platform owners | How builds reach Preview and Stable, who approves each step, the integration cycle for parallel PRs, promoting, rollback today, versioning (status, history) |
| live-testing.md | Admins, maintainers | The live test: checking sign-in, the Fabric API, OneLake and Delta against a workspace set aside for tests, with cleanup and a report |
| ideas.md | Everyone | Posting ideas, voting, and how ideas are prioritized and shipped |
| xray-fixes.md | Engineers | Each Estate X-ray finding and how to fix it |
| agent.md | Platform owners | The headless agent on Windows or Linux servers |
| architecture.md | Contributors | How the app, the engine and OneLake fit together |
| ai/README.md | Everyone | Connecting your own AI tool (Claude, Codex, Copilot, Cursor, any MCP client): setup for each, workflows, instructions to paste |
| architecture/ai-integration.md | Contributors, admins | External AI tools through MCP and a local API, with no AI in the app: architecture, permissions, security (backlog, decisions, security, compatibility, AI tool test checklist) |
| ui-guidelines.md | Contributors | Where new features go in the UI |
| ux/README.md | Contributors, product and release managers | UX and simplicity audit: scorecard, configuration, navigation, workflows, docs coverage, backlog (UX-nnn), usability release gates |
| enterprise-roadmap.md | Platform owners | Milestones for running at enterprise scale |
| local-data-engine-plan.md | Contributors, platform owners | The local data engine program FD-LDE-001 to 020: audit, implementation matrix, owners, PR plan |
| scale-and-operations-plan.md | Platform owners | Staged plan for large tables, ingestion, releases, pipelines, lineage and monitoring |
| quality/README.md | Contributors, release managers | Engineering quality baseline: scorecard, findings (QA-nnn), architecture standards, remediation plan, PR quality gate |
| git-integration-plan.md | Contributors | The Git integration program: assessment, Phase 1 (Git foundation, done), Fabric workspace Git, deployment and extensions |
| prove/plan.md | Contributors | The Prove this notebook program: what exists, the stages, and all 100 backlog items (continuation note) |
| orchestration/plan.md | Contributors | Fabric pipelines on this PC, ORCH-001 to ORCH-010: definitions, compatibility registry, local execution (continuation note) |
| workloads/plan.md | Contributors, platform owners | The workload operations program: managing workloads on PCs, agents and Fabric; stages and all 100 backlog items (continuation note) |
| optimize/plan.md | Contributors, platform owners | The Optimize program (PR-OPT-001 to 009): audit, risks, data model, PRs in order |
| finops/plan.md | Contributors, platform owners | The administration and FinOps program (FINOPS-001 to 005): audit, overlaps, PRs |
| product/enterprise-gap-analysis.md | Platform owners, contributors | The enterprise product program: what exists for each capability FD-101 to FD-140, what's missing and who owns it (backlog, decisions) |
| product/release-readiness.md | Platform owners | What must be true before an enterprise pilot, the capability matrix and the pilot checklist |
| product/PRODUCT_MANAGEMENT_PROCESS.md | Product owner, contributors | How work is chosen, ranked, built and accepted: the feature inventory, backlog, roadmap, PR-ready specs, implementation review and competitive matrix |
| why-fabric-desktop.md | Everyone | Why it exists: the Power BI model for data engineering |
| CONTRIBUTING.md | Contributors | Setting up, rules for changes, pull requests |
What is tested and what isn't#
Fabric Desktop by KratosBI is new, and it's honest about what has been proven.
Tested on every build, by automated tests on Windows and Linux, against local Delta tables:
- Notebooks, the
lakehousemodule,sparkandnotebookutilson the PC - Spark SQL translation and the SQL query engine
- The metadata framework: every load mode, checks and quarantine, surrogate keys, the four gates, compare, schema drift, replay, freshness, profiling and tests
- Startup, updates and settings
- The headless agent's queue, retries and leases
Not yet tried on a live tenant: sign-in through Microsoft's Fabric sign-in with no app registration, and Windows sign-in (version 0.15.0). Until it is, the maintainer's tenant results below are with an own app registration.
Used on the maintainer's own Fabric tenant (with an own app registration): sign-in and admin consent, listing workspaces and items, the OneLake catalog, the embedded Fabric pages (Fabric and the Monitoring hub) and their saved sign-in, and writing tables to OneLake from the PC. That is one tenant; yours may differ in settings and permissions.
Built to Fabric's documented APIs but not yet tried on a live tenant. These are tagged Preview in the app and can be hidden in Settings > Preview features:
| Preview feature | Where |
|---|---|
| Run in Fabric (notebooks and environments) | Notebook ribbon, Data mesh > Environments |
| Schedules (on this PC and in Fabric) | Notebook ribbon, Data mesh, Monitor |
| Side by side (this PC vs Fabric) and the TPC-H benchmark | Notebook ribbon |
| Prove this notebook (isolated PC vs Fabric runs, verification, receipts) | Notebook ribbon |
| New notebook from a template | Create |
| Run notebooks in their own process | Settings > Preview features |
| Cell dependencies and Run stale cells | Notebook ribbon (View and Run tabs) |
| Variable history, watches and hidden-state warnings | Notebook Variables pane (View tab) |
| Failure replay bundles (save, open and replay) | Under a cell that failed on this PC; Create > Open replay bundle |
| Notebook cell time and memory limits | Settings > Preview features (with Run notebooks in their own process on) |
| Cost | Monitor |
| Capacity status | Optimize > Cost and savings, Data mesh > Environments |
| Measured usage import | Optimize > Cost and savings |
| OneLake shortcuts | OneLake |
| OneLake access | OneLake |
| Table history and restore | Entity page |
| Table health and Optimize | Data mesh > Quality |
| Semantic model | Data mesh > Deploy |
| Tests | Data mesh > Quality |
| Estate X-ray | Home |
| Item preview | Workspaces |
| Platform (admins, landing zone, business keys, requests) | Platform, Settings, Data mesh > Surrogate keys |
| Data products for everyone | OneLake > Data products |
| Data mesh (domains from Fabric, products, links, glossary, About, the map on Home) | Data mesh > Domains, Platform > Domains, Home, entity page |
| Workloads (inventory, health, ownership, the workload page, Your workloads) | Monitor > Operations > Workloads, Home |
| Incidents (numbers, priorities, SLAs, reported incidents, your service desk) | Monitor > Operations > Incidents (incidents.md) |
Local data projects (Preview) run on your PC and don't call Fabric. Their tests use real files, DuckDB and Delta tables on every build; there are no screens for them yet (local-data.md).
Source control (Preview) is Git on your PC and doesn't call Fabric. Its tests run real Git against local repositories on every build; it hasn't yet been tried with Git Credential Manager and SSH on many PCs and servers (source-control.md).
Pipeline definitions (Preview) are read and checked on your PC and don't call Fabric. Their tests use hand-built definitions in Microsoft's documented formats; reading real exports from a tenant hasn't been tried yet (pipeline-definitions.md). The activity compatibility registry (Preview) is tested the same way; no entry has been compared against Fabric (activity-registry.md). Local pipeline runs (Preview) run control flow and child pipelines on your PC with mocks for data activities; they follow Data Factory's documentation and haven't been compared with Fabric runs (local-runs.md).
The headless agent hasn't run against a live tenant either (agent.md). Promote to Test and Prod creates lakehouses and writes through the same calls as the features above; try it on a non-production workspace first.
When a Preview feature has been tried on a tenant, its tag is removed. If you try one, say how it went in the Ideas forum or a bug report.
Need a hand?
Find a fix in troubleshooting, or share a reproducible issue with the community.
Get support ↗Improve this guide on GitHub ↗