Quick Answer
To generate an AI prompt, describe your task in one line ("grade student essays against a rubric", "write cold outreach emails for a design agency"), pick a tone and output format, and press Generate prompt. You get a complete structured prompt — role, context, task, numbered constraints, output format, and worked examples where they help — exportable as a system + user message pair for GPT, Claude, or Gemini. 3 free generations a day; unlimited with your own API key (OpenRouter, Anthropic, Gemini, or local Ollama) — the key never leaves your browser.
How It Works
Describe the task in one line — the plainer the better ("grade essays against a rubric" beats "help with grading")
Optionally add a role and audience, pick the output format the prompt should demand (prose, JSON, table, XML, bullets) and a detail level
Press Generate prompt: a complete system + user prompt streams in. 3 free runs a day, or unlimited with your own API key — it never leaves your browser
Copy the full prompt or just the system/user half, save it to the Prompt Library, refine it once more, or open it in the Optimizer or Playground
Key Facts
- One-line task description in, a complete structured prompt out
- Role/persona, tone, audience, and detail-level controls
- Output format picker: prose, JSON, table, XML, or bullet points
- Exports as system + user message pair, plain text, or a JSON messages array
- One-click second pass ("Refine once more") for more detail
- Handoffs: save to the Prompt Library, open in the Optimizer or Playground
- 3 free runs per day (resets UTC midnight); unlimited with your own API key
Frequently Asked Questions
How is this different from just asking ChatGPT to write a prompt?
You can do that — this tool IS a model call, but the surrounding structure is the value: the generator is instructed in a fixed recipe (role, context, task, constraints, output format, examples), gives you a system + user split ready to paste into an API call or another chat, and plugs straight into the Optimizer, Library, and Playground instead of dying in a chat transcript.
Does it write system prompts?
Yes. The output separates the system message (role, rules, refusal behavior) from the user message (the actual task), which is what APIs and tools like custom GPTs expect.
Can I generate prompts for image models?
For image prompts, the Image Prompt Builder is the better tool — it builds Midjourney and Stable Diffusion prompts from chips without needing an API key. This generator is for text-model prompts.
Is my task description sent anywhere?
During the free trial it goes to Google’s Gemini API to generate the prompt (never logged here). With your own key it goes straight from your browser to your provider — this site’s servers never see it.