Workshop Desktop

General AI Agent for your Computer

Research and analyze data, build apps, create tools, prototype ideas, and glimpse AGI with the privacy and power of your desktop.

Unlimited free usage with local AI

See in action

Scroll to explore demos, or tap to jump.

interactively research my data tables in

BigQuery. I want to tie together the relevant quickbooks export, customer, and billing data tables to determine CAC and LTV by weekly cohorts.

I'll connect to your BigQuery data warehouse and explore the available tables.

Found 3 relevant tables. Let me inspect their schemas to understand the data structure.

The quickbooks_export table contains marketing spend by channel. Let me sample recent data.

Now I'll calculate Customer Acquisition Cost (CAC) by weekly cohort by joining marketing spend with new customer signups.

Excellent progress - CAC is trending down. Now let me calculate 6-month LTV for each cohort.

Here's how your unit economics have evolved across cohorts:

CAC & LTV by Weekly Cohort

And the LTV/CAC ratio - the key metric for sustainable growth:

LTV/CAC Ratio by Cohort

Key Insights:

CAC decreased 23% over the 8-week period ($60 → $46)
Week 3 cohort (Oct 21) has highest LTV at $247
LTV/CAC ratio improved from 3.0x to 4.0x
Recommendation: Google Ads shows best ROI - consider increasing budget allocation

Ask follow-up questions or request deeper analysis...

I need to prototype the different user flows in our product's dashboard based on subscription tiers. Can you create a demo mode with a plan selector that toggles between Pro and Max tiers? I want to show stakeholders how features and limits differ.

I'll create a demo mode that lets you toggle between Pro and Max tiers. Let me first explore your codebase to understand the existing structure.

Found the existing subscription store. Let me read it to understand the current structure.

I see you have a basic plan config. I'll create a dedicated subscription store for the demo mode with detailed tier features.

Now I'll create the plan selector toggle component.

Next, I'll build the feature grid that shows enabled/locked features based on the selected tier.

I'll add usage meters that show how limits change between tiers.

Finally, I'll wire everything together in the dashboard page.

All components are ready. Let me start the development server so you can test it.

Your demo mode is ready! Open http://localhost:5173 in your browser to test it.

Use the plan selector in the dashboard header to toggle between Pro and Max tiers. You'll see:

Pro tier: 4 enabled features, 3 locked with upgrade hints, capped usage limits
Max tier: All 7 features unlocked, unlimited usage displayed

Ask to customize features or add more tiers...

Help me set up my Mac to build a native iOS app. I want to create a simple task tracker app in Swift with SwiftUI.

I'll help you build a TaskTracker iOS app with SwiftUI. Let me first check your development environment and then set up the project.

Xcode is installed. Now I'll install XcodeGen to generate the project from a simple YAML configuration.

Now I'll create the project configuration file that defines your app's structure.

Project generated. Now I'll write the SwiftUI app code starting with the main app entry point.

All Swift files are ready. Let me find an available simulator to run your app.

I'll use iPhone 15 Pro. Building the app now...

Your TaskTracker app is now running in the iPhone 15 Pro simulator! I've created a complete SwiftUI task tracker with:

Task list with checkable items and swipe-to-delete
Add task sheet with form validation
State management using ObservableObject pattern
Modern SwiftUI with NavigationStack and toolbars

Would you like me to add persistence with SwiftData, task categories, or due dates with notifications?

Ask to add features like persistence, categories, or notifications...

I need to add the new PostHog events from our product codebase to our dbt analytics pipeline. The product team just added tracking for the checkout flow—can you analyze what events they added and create the dbt models?

I'll help you create dbt models for the new checkout events. Let me first search your product codebase to understand what PostHog events were added.

Found 3 checkout-related events. Let me read the analytics file to understand the event properties.

I can see the three events: checkout_started, payment_method_selected, and order_completed. Now let me check your existing dbt structure to maintain consistency.

I'll follow your existing naming conventions. First, I'll create a staging model to transform the raw PostHog events.

Now I'll create a mart model that calculates the checkout funnel metrics—conversion rates and revenue by user.

Let me add the schema tests to ensure data quality.

Now let me build and test the models to verify everything works.

Summary:

Created staging model stg_posthog__checkout_events—transforms raw PostHog events
Created mart model fct_checkout_funnel—user-level funnel metrics with conversion rates
Added 6 data tests—uniqueness, not-null, and accepted values constraints
All 8 tests passed—models built and validated successfully

Next steps: You can query fct_checkout_funnel directly in your BI tool, or I can help you create a checkout analytics dashboard.

Ask about the funnel metrics or request a dashboard...

See in action

Scroll to explore demos, or click to jump.

interactively research my data tables in

BigQuery. I want to tie together the relevant quickbooks export, customer, and billing data tables to determine CAC and LTV by weekly cohorts.

I'll connect to your BigQuery data warehouse and explore the available tables.

Found 3 relevant tables. Let me inspect their schemas to understand the data structure.

The quickbooks_export table contains marketing spend by channel. Let me sample recent data.

Now I'll calculate Customer Acquisition Cost (CAC) by weekly cohort by joining marketing spend with new customer signups.

Excellent progress - CAC is trending down. Now let me calculate 6-month LTV for each cohort.

Here's how your unit economics have evolved across cohorts:

CAC & LTV by Weekly Cohort

And the LTV/CAC ratio - the key metric for sustainable growth:

LTV/CAC Ratio by Cohort

Key Insights:

CAC decreased 23% over the 8-week period ($60 → $46)
Week 3 cohort (Oct 21) has highest LTV at $247
LTV/CAC ratio improved from 3.0x to 4.0x
Recommendation: Google Ads shows best ROI - consider increasing budget allocation

Ask follow-up questions or request deeper analysis...

I'll create a demo mode that lets you toggle between Pro and Max tiers. Let me first explore your codebase to understand the existing structure.

Found the existing subscription store. Let me read it to understand the current structure.

I see you have a basic plan config. I'll create a dedicated subscription store for the demo mode with detailed tier features.

Now I'll create the plan selector toggle component.

Next, I'll build the feature grid that shows enabled/locked features based on the selected tier.

I'll add usage meters that show how limits change between tiers.

Finally, I'll wire everything together in the dashboard page.

All components are ready. Let me start the development server so you can test it.

Your demo mode is ready! Open http://localhost:5173 in your browser to test it.

Use the plan selector in the dashboard header to toggle between Pro and Max tiers. You'll see:

Pro tier: 4 enabled features, 3 locked with upgrade hints, capped usage limits
Max tier: All 7 features unlocked, unlimited usage displayed

Ask to customize features or add more tiers...

Help me set up my Mac to build a native iOS app. I want to create a simple task tracker app in Swift with SwiftUI.

I'll help you build a TaskTracker iOS app with SwiftUI. Let me first check your development environment and then set up the project.

Xcode is installed. Now I'll install XcodeGen to generate the project from a simple YAML configuration.

Now I'll create the project configuration file that defines your app's structure.

Project generated. Now I'll write the SwiftUI app code starting with the main app entry point.

All Swift files are ready. Let me find an available simulator to run your app.

I'll use iPhone 15 Pro. Building the app now...

Your TaskTracker app is now running in the iPhone 15 Pro simulator! I've created a complete SwiftUI task tracker with:

Task list with checkable items and swipe-to-delete
Add task sheet with form validation
State management using ObservableObject pattern
Modern SwiftUI with NavigationStack and toolbars

Would you like me to add persistence with SwiftData, task categories, or due dates with notifications?

Ask to add features like persistence, categories, or notifications...

I'll help you create dbt models for the new checkout events. Let me first search your product codebase to understand what PostHog events were added.

Found 3 checkout-related events. Let me read the analytics file to understand the event properties.

I can see the three events: checkout_started, payment_method_selected, and order_completed. Now let me check your existing dbt structure to maintain consistency.

I'll follow your existing naming conventions. First, I'll create a staging model to transform the raw PostHog events.

Now I'll create a mart model that calculates the checkout funnel metrics—conversion rates and revenue by user.

Let me add the schema tests to ensure data quality.

Now let me build and test the models to verify everything works.

Summary:

Created staging model stg_posthog__checkout_events—transforms raw PostHog events
Created mart model fct_checkout_funnel—user-level funnel metrics with conversion rates
Added 6 data tests—uniqueness, not-null, and accepted values constraints
All 8 tests passed—models built and validated successfully

Next steps: You can query fct_checkout_funnel directly in your BI tool, or I can help you create a checkout analytics dashboard.

Ask about the funnel metrics or request a dashboard...

Create the impossible

Workshop bundles the features you need to create, and the tools AI needs to help you do so safely.

78% on SWE-Bench Verified

Industry-leading performance on real-world engineering challenges.

Credentials Management

Securely store credentials that utilize your computer's built-in credentials management.

Use MCP Servers

Install MCP servers with one-click and use them from the agent.

Build MCP Servers

Need an MCP server? Workshop can autonomously build and test MCP servers for you.

Build for any tech stack

Get up and running in 60 seconds on Windows, Mac, and Linux.

Built-in Templates

Sync your projects from Workshop Cloud to continue building on your desktop.

Edit the code

Conversations and project files are stored locally on your machine.

One-click install

Get up and running in 60 seconds on Windows, Mac, and Linux.

Web Sync

Sync your projects from Workshop Cloud to continue building on your desktop.

Privacy Mode

Conversations and project files are stored locally on your machine.

Built-in, local code interpreter

Chat with data on your machine or in the cloud using Workshop Desktop's built-in code interpreter. Ask the agent to run queries, visualize data, and write + execute ad hoc scripts.

Work with data on your computer

Whereas Workshop Cloud is optimized to take you from prompt-to-data-app, Workshop Desktop is like a software engineer that lives on your computer and can help you with any technical task.

"Oh my god! You guys are the best so far in my experience. I've tried lovable, bolt, databutton, Rork, Base 44, and another one that YC is sponsoring, but it sucks. So believe me when I say you guys are the best! Keep up the great work!"

Oh my god! You guys are the best so far in my experience. I've tried lovable, bolt, databutton, Rork, Base 44, and another one that YC is sponsoring, but it sucks. So believe me when I say you guys are the best! Keep up the great work!

Workshop seems to be, so far, exactly what i've been looking for. i was able to use it to prototype my dream note taking app in one morning. It's very impressive.

Out of all of them i tried. workshop can build any app in any language. Its also great at reverse engineering android apps and making alternative versions with more functions than the original

Just found workshop last night. Absolutely blown away so far. Ive tried to build this MVP on a number of tools and didnt even get close, if I did, design was not usable

Yo Workshop is Filthy. It one shot a Soundcloud to wav downloader that I'm using to build a DJ set for my friends wedding

[Workshop] is becoming part of my team. From 5-minute tasks to 40-hour planning, development, and execution. It's starting to give me superpowers

Ready to start creating?

Start creating with Workshop Cloud