For agents & the people driving them

Point your AI at QuizGen.

QuizGen is a remote MCP server. Connect it once, authorize your account in the browser, and your assistant can create hosted quizzes, read responses, and run follow-up workflows in every conversation.

Connect the MCP server

Use the same remote endpoint in ChatGPT, Claude, Codex, Cursor, VS Code, or any client that supports streamable HTTP MCP. Opening the setup page does not connect a plain chat by itself—the client must add the MCP server.

Open setup for my client

$ https://quizgen.dev/mcp

Codex

codex mcp add quizgen --url https://quizgen.dev/mcp --oauth-resource https://quizgen.dev/mcp
codex mcp login quizgen

Claude Code

claude mcp add --transport http --scope user quizgen https://quizgen.dev/mcp
# Then open /mcp in Claude Code and authorize QuizGen.

Use outcome-first prompts

You do not need to name tools or describe the JSON schema. Ask for the whole result and let the connected agent choose the tools.

Create and publish a five-question client intake form asking for name, budget, timeline, goals, and what they have already tried. Give me the live link.

Find my client intake quiz, summarize completed responses, show the budget breakdown, and rank the strongest leads with your reasoning.

Email my onboarding quiz to alex@example.com. Before sending, show me the recipient and message for confirmation; after I confirm, schedule one reminder in three days for non-completers.

Optional: add the QuizGen skill

MCP gives your assistant access to QuizGen. The optional skill teaches it stronger quiz design, recovery behavior, and the full API schema. It complements MCP; it does not replace the connection.

Download skill (.zip)Claude or ChatGPT: upload in Customize → Skills

$ npx skills add Tariqlearnstocode/quizgen-skills

Codex, Claude Code, Cursor, or another terminal agent: run the command above.

Raw HTTP fallback for executable agents

Use this only when the agent can actually execute HTTPS requests. No Authorization header creates a 72-hour sandbox with a private claim link. API keys belong in secret configuration—not prompts.

curl -X POST https://quizgen.dev/api/v1/quizzes \
  -H "Content-Type: application/json" \
  --data '{ "schema_version": 1, "title": "Coffee or tea?",
    "sections": [ { "key": "main", "questions": [
      { "id": "pick", "type": "choice", "prompt": "Coffee or tea?",
        "options": [ { "value": "coffee", "label": "Coffee" },
                     { "value": "tea", "label": "Tea" } ] } ] } ] }'
# → { "url": "https://quizgen.dev/q/x7km2p", "sandbox": true, ... }

Everything is also markdown

/mcpcanonical MCP endpoint and client setup
/llms.txtraw HTTP API reference for executable agents
/index.mdthe landing page
/pricing.mdplans and limits
/blog.mdguides — scoring, webhooks, RSVPs, MCP setup, and more
/examplesexample quizzes with the prompts that make them