Start by describing a real, tightly-scoped problem or desired integration in an Amp thread.
Examples: "Integrate Groq API into Libre Chat," "Investigate Groq provider conversation flow issue," "Fix OpenRouter model dropdown population issue."
Include any error output, screenshots, and summary of symptoms or gaps.
Ask for a research-backed plan from the Oracle before requesting code changes, e.g. "Use the Oracle to research… then create an implementation plan."
Amp responds with detailed TODOs and step-by-step plans specific to the stack, based on source code investigation and external API documentation.
When blocked, paste exact errors or system output from your runtime (build, test, or app logs).
Amp traces back the problem using source search ("search for OpenRouter-related code," "find where provider value is being passed") and iteratively proposes specific code or config adjustments.
After each code change, Amp will automatically pompt itself to test: "Now let me test the fix by running X," and report the resulting output or error.
Amp examines build output, backend logs (e.g. podman, npm errors), and guides any missing dependency installs or required container restarts.
For frontend changes, Amp will validate the UI/UX changes in the browser. Amp can also use an MCP such as Playwright to automate browser testing.
Amp ties backend config, server environment, and frontend build together for end-to-end fixes.
If the first fix doesn't work, update the thread with current symptoms/logs, and repeat with Amp until resolved.
The workflow is highly interactive and dialogue-driven, treating Amp as your pair-programming partner for both research and incremental implementation.
- Tightly-scoped problems: Focus on specific, actionable issues rather than broad requirements
- Research-first approach: For comprehensive changes, always start with investigation and planning before implementation
- Error-driven debugging: Use actual logs and outputs to guide problem-solving
- End-to-end validation: Verify fixes work from backend to frontend
- Incremental iteration: Test after each change and adjust based on results
✅ Good: "Integrate Groq API into Libre Chat - getting 401 errors on authentication"
✅ Good: "OpenRouter model dropdown not populating - see attached screenshot"
❌ Avoid: "Make the app better" or "Add AI features"
❌ Avoid: "Fix all the bugs" or "Optimize performance"
- Specific problem described with context
- Research plan requested and received
- Errors/logs shared for debugging
- Each fix tested and validated
- UI changes verified in browser
- Iterative improvement cycle maintained