Guide: Building and Shipping
This guide is for Engineers, Tech Leads, and QA. It covers three stages: Engineering (planning), Development (building), and QA (testing). By this point, your project has confirmed requirements, resolved blockers, and (optionally) completed designs.
Engineering: Plan the Architecture
The Engineering stage is about creating a technical blueprint before anyone writes code.
Open the project and start chatting. The AI has full context — the PRD, requirements, acceptance criteria, and designs. But more importantly, it researches your actual codebase through the Knowledge Layer. It knows your existing patterns, data models, APIs, and conventions.
This means the AI can help you:
- Identify which files and modules need to change
- Spot existing patterns to follow (or intentionally deviate from)
- Define API contracts and data model changes
- Flag technical risks or dependencies
The output is an Engineering document — tagged as engineering in the project — that serves as the implementation plan.
When Is Planning Done?
A requirement's dev status moves to Planned once the engineering plan covers it. The Engineering stage is complete when all confirmed requirements are at least Planned.
Development: Build It
This is where plans become code. Developers work through requirements one by one, updating their status as they go:
Not Started → Planned → Developed → Tested
Using ProdE with Your Coding Agent
The real power of the Development stage is the MCP integration. Connect ProdE to your coding tool (Cursor, VS Code, Windsurf, or Claude Code), and your coding agent gets full project context:
- The PRD and all project documents
- Every requirement with its acceptance criteria
- The engineering plan
- Blocker resolutions and design decisions
This means when you tell your coding agent "implement REQ-5," it already knows what REQ-5 is, what the acceptance criteria are, what the engineering plan says, and how your existing codebase works.
See Connecting to Your Coding Agents for setup instructions.
Updating Requirement Status
As you work, update each requirement's dev status:
| Status | When to Set It |
|---|---|
| Planned | Engineering plan is in place for this requirement |
| Developed | Code is written and ready for testing |
| Tested | Implementation has been verified by QA |
You can update status through the ProdE UI or through MCP tools in your coding agent.
Logging Bugs
If you find issues during development, create Bugs in the project. Each bug gets a unique ID (BUG-1, BUG-2) and tracks through:
Open → In Progress → Closed
→ Not a Bug
Keeping bugs inside the project means they're visible alongside requirements and progress — no need to context-switch to a separate tool.
QA: Verify It Works
The final stage. QA reviews each requirement's acceptance criteria and verifies the implementation meets them.
The AI can help plan testing strategies based on acceptance criteria. For a requirement like "Users can sign in using their Google account," the AI might suggest:
- Verify the Google login button appears on the login page
- Verify a new account is created on first Google login
- Verify an existing user with the same email is linked correctly
- Verify login fails gracefully if the user denies Google permissions
- Verify the session is created with the correct expiry
Once a requirement passes testing, its dev status moves to Tested. The QA stage — and the project — is complete when all confirmed requirements reach Tested status.
Tracking Progress Along the Way
At any point during Engineering, Development, or QA, you can check the Pipeline to see where things stand. It shows how many requirements are at each stage and where bottlenecks are forming.
See Guide: Tracking Delivery for details.
Next: Understand how to track overall delivery readiness — Guide: Tracking Delivery.