Model, query, and map your data — local-first, in your browser or on your desktop.
Open the web studio and start modelling in seconds — no install, no sign-up to try. The desktop app adds live connections to your own databases and Git source control.
macOS · Windows · Linux · Web — free up to 15 tables, no email required
The web studio reaches your model. The desktop app reaches your infrastructure.
Both are local-first and share the same modelling core, and one Pro or Team subscription covers both. The desktop app adds live connections to your own databases, built-in Git source control, and fully offline work.
One shared schema model
Edit once. Everything follows.
Every workspace reads from the same model — design a table and it's instantly in your SQL autocomplete, your DDL output, your lineage graph, and your glossary.
Schema canvas
Design tables, views, and custom types on a drag-and-drop canvas. Or write it as code: the DBML Workspace renders your canvas as live DBML — free on every tier, with exports that keep your full diagram layout in the file.
SQL workspace
A multi-tab editor with autocomplete fed live from your canvas. Run against an embedded DuckDB sandbox seeded from your schema, a local .duckdb file, or MotherDuck — with result charts, EXPLAIN plans, and query history. The desktop app adds your own connected databases.
Lineage & impact analysis
Table- and column-level lineage across sources, jobs, and consumers — derived automatically from your views' SQL definitions. See the full blast radius before you drop a column.
Source-to-target mapping
Column-level mappings with transformation nodes — joins, filters, aggregates, custom SQL. Exports executable INSERT or CTAS statements in your target dialect, not just documentation.
DDL in 13 dialects
PostgreSQL, Snowflake, BigQuery, SQL Server, Databricks, DuckDB, Redshift, ClickHouse, and more. Full CREATE scripts, ALTER migrations between versions, and realistic seed data.
Live database drivers
Eight native drivers in the desktop app, plus DuckDB WASM and MotherDuck everywhere. Introspect production schemas onto the canvas, detect drift against what's deployed, and generate the migration that closes the gap — including databases on your own private network.
Trust, verifiable
Built to pass your security review.
Most data tools ask you to trust their cloud. Wurk.Flow doesn't have one to trust — your project data never leaves your machine, a file on your disk on desktop and kept in the browser on web, and every byte of network traffic it produces is documented for your team to audit.
Read the full network disclosureLocal-first by design
Schemas, credentials, query results, and project data stay on your device — a file on your disk on desktop, and in the browser on web. Nothing is sent to us unless you opt in: a Team Co-Op session syncs canvas state through our relay, or one you self-host. On desktop the core workflow also runs with no network connection at all.
Every outbound call documented
No usage analytics, no tracking, on the site or in the app. The desktop app makes three baseline network calls — license validation, click-only update checks, and scrubbed crash reports. The web studio makes one baseline call: the subscription check for a signed-in account. Signed out, it makes none. All listed publicly for your security review.
You keep what you pay for
Annual desktop licenses include perpetual fallback: stop renewing and the version you paid for keeps working — we renew that license for free, automatically, for as long as you use it. Everything your team built stays yours regardless: projects are plain-text .dsm files you can diff in your own Git, and DBML export is free on every tier — including Free.
Your AI keys, your traffic
AI features are bring-your-own-key: seven providers on desktop — OpenAI, Anthropic, Azure, Bedrock, Groq, Mistral, and Ollama — and OpenAI, Anthropic, Groq, Mistral, or local Ollama in the browser. Requests go straight from the app to your provider; nothing routes through our infrastructure. And with the desktop MCP server (Pro), Claude Code, Cursor, or Claude Desktop can read and edit the model you have open — over loopback, with a token that never leaves your machine.
Scoped on purpose
Wurk.Flow is a modeling IDE, not a platform that claims to do everything. It is not a pipeline orchestrator, an org-scale data catalog, a BI tool, or a database admin console — and it isn't yet SOC 2, HIPAA, or FedRAMP certified. We'd rather you find out it does more than you expected, not less.
Where Wurk.Flow fits
Serious capability, without the procurement cycle.
Teams today stitch together a diagramming web app, a query IDE, and a spreadsheet of mappings — then spend hours keeping them in sync. Wurk.Flow replaces all three.
Lightweight web tools
dbdiagram, DrawSQL
$100–200/yr for diagrams and DDL scripts alone — no mapping layer, no lineage, no query IDE. The database is something you export to, not work in. And stock DBML drops your canvas layout on export; ours keeps it in the file.
The gap
Wurk.Flow
Lineage, source-to-target mapping, and a live SQL workspace in one local-first IDE. $99/yr, self-serve — no quote, no procurement cycle. macOS, Windows, Linux, or the browser.
Enterprise modelers
erwin Data Modeler
$3–4K+/seat/yr list · Windows-only · quote-based enterprise procurement.
Free tier. No email. No expiry.
Up to 15 tables and views, 3 DDL dialects, the DBML Workspace, and the schema canvas — yours indefinitely, in the browser or on the desktop. Upgrade when your team outgrows it, or don't.
- No credit card
- No sign-up to try
- Free tier never expires
Rolling it out across a team?
Wurk.Haus runs hands-on engagements — implementation, training, pipeline integration, and data-contract authoring — led by the engineer who builds the product.