C3 AI Documentation Home

Build, Share, and Deploy Your First Application

In this tutorial, you build an enterprise application and deploy it to production. The end result is an example application that is backed by a real data pipeline. You build the application from a single prompt, review and refine it, check the data pipeline and model the agent built, then share and deploy it.

The example is a spare-parts inventory application. The application tracks how much of each part is in stock and flags any part running low, so you can reorder it before it runs out. The pages show where those shortages sit across your locations.

Before you begin

Confirm all three of the following before you start:

  • You can open C3 Code in your C3 Agentic AI Platform cluster.
  • Your quota allows you to create at least one application and one workspace. The My Usage and Cost page shows both counts, and an administrator can raise either count. See Request Access and Resources.

Step 1: Describe the application you want

The C3 Code home page opens on a single prompt input:

  1. Describe the application. Name the records it stores, the logic it runs, and the pages you expect. To use a requirements document you already have, attach the document instead of typing the description. You can also attach data files, or ask C3 Code to generate sample data to start quickly.
  2. Select the arrow icon to submit. In the modal that opens, enter your application name and select Create.

One prompt builds the records and the pages and creates the data. Name what to store, the logic to run, and the pages you want, then ask the agent to generate sample data to start with:

Text
Build an inventory monitoring application called PartsInventory. Create the records, some
starting data, and all the pages in one pass.

Store an Item with a part number, description, product family, unit cost, and ABC class. Store
an Inventory Position with an item, location, region, supplier, units on hand, units in
transit, average weekly demand, and lead time in days.

Work out days of supply for each position, and give each one a status: Stockout at 0 units,
At Risk under 14 days, Healthy, or Excess over 90 days.

Build a Home page with summary cards, a chart of healthy and at-risk value by region, and a
chart of positions by status, and a Reorder page listing everything that needs reordering.

Also generate sample data representing industrial spare parts inventory, using the fields
above:

SKU, Description, ProductFamily, AbcClass, Location, Region, Supplier, OnHandUnits,
InTransitUnits, AvgWeeklyDemand, LeadTimeDays, UnitCost

Use 12 distinct SKUs across 4 distribution centers: DC-EAST in the Southeast, DC-NORTH in the
Midwest, DC-WEST in the West, and DC-CENTRAL in the Southwest. Vary the numbers so the data
tells a story: a few rows with OnHandUnits of 0, several rows where on-hand covers less than
two weeks of demand, and at least one row holding thousands of units of a part costing a few
dollars. Use AbcClass values of A, B, or C, and lead times between 7 and 45 days. Integrate
the data through Data Fusion Pipeline.

The C3 Code home page with the opening of the build prompt entered and no file attached

C3 Code creates the application and sends your prompt to the agent. The workspace opens on the New Tab page, which lists every tool you can open.

While the agent works, the chat panel reports each step as it happens, including every file the agent writes and the number of lines it added. From the example prompt, the agent's work covers the following:

  • Create the Item and InventoryPosition records, with the days-of-supply calculation and the status of each position.
  • Generate the sample data for 12 SKUs across 4 distribution centers and load it through a Data Fusion pipeline.
  • Build the Home page with summary cards, a chart of healthy and at-risk value by region, and a chart of positions by status, and the Reorder page listing every position that needs reordering.
  • Add Home and Reorder to the navigation.

The workspace chat panel reporting the agent's progress, with each file it wrote and the lines it added, beside the Preview tab

When the agent finishes, it summarizes what it built and lists decisions you may want to change.

The controls you use for the rest of this tutorial sit in two places. The workspace header carries the Deploy and Share buttons, and a Connect to GitHub button that becomes the Git status control after you link a repository. The tab strip, the row of tabs beside the chat panel, holds your tools. Select Open Preview on the New Tab page to see your application, and select + to open Code or another tool as a tab.

Building an application of this size takes several minutes. You can type a follow-up prompt while the agent works.

Step 2: Review and refine

When the agent finishes, the preview shows the running application on the starting data the agent generated. Ensure the following:

  1. The pages and the navigation match what you asked for.
  2. Every page renders with real data, rather than an empty screen.
  3. The tables and charts carry the field names you asked for.

Generated inventory application running in the Preview tab, with summary cards, a chart of value by region, and a chart of positions by status

Each follow-up prompt builds on what already exists, so you never start over. Enter one change in the chat panel, such as a new field, a filter, or a page, then review the result in Preview.

For writing a first prompt and refining the result, see Start from Scratch.

Step 3: Check the data pipeline and model

The agent did more than fill the screens with numbers. It generated the sample data you asked for and loaded it through a real pipeline built in C3 Data Fusion, the same kind of pipeline that carries data from a file or from a system your company runs. Open both to see what the agent built for you.

Select + next to your open tabs, then Data Fusion. On its Data Integration tab, the pipeline runs from the generated file into the records themselves. It passes through your file source system and one transform for each kind of record:

The Data Integration canvas showing the pipeline from the generated file through two transforms into the Item and InventoryPosition records

The Object Model tab shows the data model, so you can confirm the agent built real records with the fields and logic you asked for:

The Object Model tab showing record cards with their fields, types, and methods

Together, these confirm the application holds real records behind the pages, not placeholder screens. When you are ready to run it on your own data, ask the agent to connect the system that holds your records, and enter the sign-in details in the Data Connector Wizard. You can also attach a file to a prompt. To find relevant tutorials, see Use Configuration Tools.

Step 4: Share the application with a colleague

Sharing works at two levels. Application collaborators contribute to the application, and preview recipients only view a running workspace.

To add an application collaborator:

  1. Select Applications, then select your application.
  2. Select Collaborators > Add collaborator.
  3. Enter your colleague's email address. The picker takes an existing user, or an address you type.
  4. Set the role. User creates workspaces and contributes. Admin additionally manages all workspaces in the application, deploys, and adds other administrators.
  5. Select the add button to confirm.

Add Collaborators dialog with the User and Admin role descriptions

Your colleague appears in the Collaborators table with the role you assigned.

To send a link to the running workspace instead:

  1. Select Share in the workspace header.
  2. Copy the preview URL, or enter an email address and select Share.

Share Preview dialog showing the preview URL, the Add collaborators search, and the collaborator list

The person you add becomes a workspace user who can view but not contribute. The link serves the running workspace, so the link stops working while that workspace is stopped.

For a detailed tutorial on adding people and sharing a preview link, see Share with Collaborators.

Step 5: Deploy to staging

A deployment runs independently of your workspace, so your colleagues reach the application whether or not your workspace is running. Staging is where you and your colleagues iterate on features without affecting your users. Production is a separate environment that your users reach, and it holds the most stable copy of your application. For a new change, a best practice is to first deploy to staging, test and interact with the feature in Preview, then promote the deployment to the production environment.

  1. Select Deploy in the workspace header. The Deploy Application popover opens with Staging already selected.
  2. Confirm Staging is selected.
  3. Select Deploy to Staging, or Redeploy to Staging when staging already has a deployment. The button reads Validating while C3 Code checks your application.
  4. In the Confirm Deployment dialog, select Continue.

Deploy Application popover with the Staging tab selected, No active deployment, and Deploy to Staging

When the check finds problems, the Package Issues Detected modal opens first and groups those problems by file. Select Fix with C3 AI Assistant to pass the problems to the agent. Errors block the deploy until the agent or you fix them. Warnings do not block the deploy, but they can leave the deployed application in a degraded state, so fix them first when you can. To deploy with warnings anyway, see Deploy to Staging and Production.

The Confirm Deployment dialog names the environment you are deploying to. When a deployment already runs in that environment, continuing updates it:

Confirm Deployment dialog asking you to confirm the deploy to the Staging environment, with Cancel and Continue

After you select Continue, you can monitor the deployment's progress in the popover. For details on each step, see Deploy to Staging and Production.

Deploy Application popover showing a first staging deployment at step 1 of 6, Packaging application, with the time remaining and the first-deployment note

C3 Code packages the files exactly as those files stand in your workspace. You can keep working while the deploy runs, and changes you make afterward stay in your workspace.

When you close the popover, a banner at the top of the workspace shows the current step. Select View progress in the banner to reopen the popover. Once the deployment finishes, select Launch Application to open the running application.

Step 6: Deploy to production

Deploying to production follows the same steps as deploying to staging.

  1. Confirm staging runs the version you want your users to have.
  2. Select Deploy in the workspace header.
  3. Select Production.
  4. Select Deploy to Production, then clear any package issues the check reports.
  5. In the Confirm Deployment dialog, select Continue.

Each application has one staging deployment and one production deployment, so deploying again to the same deployment replaces the running version. Confirm which deployment you selected before continuing.

For further details, see Deploy to Staging and Production.

Check the result

Select Deployments in the navigation pane. The list shows one row per deployment, with the application, the version, the environment, the status, the hibernation schedule, and the last deployed time.

Deployments page listing the staging and production deployments for the application

Confirm all three outcomes:

  • Your application appears with a Staging row and a Production row.
  • Each row carries a version stamp.
  • Your colleague appears in the Collaborators tab of the application.

When a deployment reports a status other than running, open the deployment to read the reason before deploying again.

Where to go next

These pages take each stage of the quick start further:

Was this page helpful?