Your first program

With Foundry running locally, seed the AcmeCorp reference dataset to explore the full delivery pipeline. AcmeCorp is a realistic demo program with 118 requirements, 8 skills, and 7 workstreams spanning two migration corridors (Magento to Salesforce B2B Commerce, Magento to BigCommerce B2B).

Seed the AcmeCorp demo

1. Get your organization ID

   After signing into Foundry and creating a Clerk organization, you need the organization ID. Find it in:

   • The Clerk dashboard under Organizations
   • Or the browser URL when viewing your org settings

   The ID looks like org_2abc123def456.

2. Run the seed command

   With Convex dev running, open a new terminal and run:

   bunx convex run seed:seedAcmeCorp '{"orgId": "org_YOUR_ORG_ID"}'

   Replace org_YOUR_ORG_ID with your actual Clerk organization ID.

   This creates:

   • 1 program (AcmeCorp Migration)
   • 7 workstreams (Catalog, Orders, Customers, Integrations, etc.)
   • 118 requirements across all workstreams
   • 8 skills (migration-specific capabilities)

   Note

   The seed operation runs as a Convex action and takes a few seconds to complete. Check the Convex dev terminal for confirmation.

3. Open the program

   Navigate to http://localhost:3000. The AcmeCorp program now appears on your Programs page. Click into it.

Exploring the program

Once inside the AcmeCorp program, the main navigation surfaces are:

Mission Control

The program dashboard. Shows an overview of delivery status, recent activity, and health scores. This is where you start each session.

Discovery

The requirements intake pipeline. Documents uploaded here are analyzed by Claude to extract structured requirements. Each finding maps to workstreams and skills.

To run your first discovery pass:

1. Navigate to the Discovery tab within the program.
2. Upload a document (PDF, DOCX, or plain text) describing requirements — or use the seeded data to explore existing findings.
3. Trigger an analysis. Claude processes the document and extracts structured findings.
4. Review the findings queue. Each finding can be accepted, rejected, or refined.

Tip

Discovery uses Claude Opus 4.6 for document analysis. Make sure your Anthropic API key is set in the Convex Dashboard environment variables.

Tasks

The task board shows decomposed work items generated from requirements. Tasks are scoped to specific skills and can be assigned to sandboxed AI agents for execution.

Key features:

• Task decomposition — requirements break into granular tasks with verification criteria
• Subtask generation — tasks further decompose into implementation subtasks
• Sandbox execution — provision an ephemeral Docker container with Claude Code for a task

Workstreams

Workstreams group related requirements into delivery tracks. The AcmeCorp demo includes 7 workstreams:

• Catalog Migration
• Order Management
• Customer Data
• Integrations
• Frontend/UX
• Infrastructure
• Data Migration

Each workstream shows its requirements, tasks, and completion progress.

Skills

Skills represent domain-specific capabilities that Foundry agents use when executing tasks. The AcmeCorp demo seeds 8 skills covering areas like data migration, API integration, and frontend development.

Skills include:

• Detailed instructions for the AI agent
• Design context and constraints
• Repository mappings

Risks

AI-assessed risk evaluations for the program. Risk scoring considers requirement complexity, dependency chains, and implementation status.

Next steps

• Review the environment variables to enable additional integrations (GitHub App, Atlassian, sandbox worker)
• Read the architecture overview to understand how requirements flow through the system
• Connect a GitHub repository to enable source control features
• Try running a sandbox execution on a task to see the AI coding pipeline in action