The interview before the interview
Dry Run is an interview-practice app that turns a job posting and optional résumé into a role-specific rehearsal. A candidate answers naturally; an agent uses explicit WebMCP tools to set context, retrieve a brief, start the interview, submit answers, and return a grounded verdict.
Why WebMCP
Interview preparation has useful context trapped in page UI. Dry Run exposes the core workflow through structured tools, so people and agents share the same capability layer rather than relying on brittle UI automation. The human controls the conversation while the agent handles the structured setup, progression, and scoring flow.
What people and agents can do together
- Set a role and optional résumé context.
- Generate a grounded brief and eight targeted questions.
- Run a realistic question-and-answer practice session.
- Submit answers and receive a structured final verdict.
Implementation
Dry Run is a Svelte 5 and Vite single-page app deployed on Netlify. It registers six WebMCP tools with document.modelContext.registerTool: set_posting, set_resume, get_brief, start_interview, submit_answer, and get_verdict. Human UI interactions and tool calls share the same session functions for parity.
The app uses OpenAI gpt-5.6-luna through Netlify AI Gateway for structured role analysis and answer feedback. It has been tested in ChatGPT's in-app browser with GPT-5.6 Sol and Terra.
Testing
Open https://dryrun.nryn.dev/ in ChatGPT's in-app browser. Use the WebMCP tools to set a job posting and optional résumé, retrieve the brief, start the interview, submit an answer, and retrieve the final verdict.
Built With
- gpt-5.6-luna
- netlify
- netlify-ai-gateway
- openai
- pdfjs-dist
- svelte-5
- vite
- webmcp