Help · Guides · PromptGen

PromptGen — production prompts in 10 seconds.

App: /apps/promptgen/ Cost: 2 credits per generate Time: 3 min · ships .md export

PromptGen takes a task description and turns it into a production-quality system prompt + user prompt + usage guidance — tuned for the model you're targeting. Skip the trial-and-error round of prompt drafting.

Open PromptGen

Go to /apps/promptgen/. If you're on Plus / Studio / Crafted, you're already authorized — your account email is auto-detected from your portal session. If not, sign in at /portal/ first.

Plus tier costs 2 credits per generate. Plus gets 75 credits/month — that's ~37 generates included in the base $29/mo.

Describe the task

In the What should the prompt do? field, write 1–3 sentences describing the job. Be specific about inputs and outputs.

Good: "Classify customer support emails into one of: billing, technical, feature_request, complaint, other. Return JSON with the category and a one-line summary."

Too vague: "Help with support emails."

Tip — include the constraints upfront. Mention output format (JSON / markdown / plain text), tone (formal / casual), and length limits ("under 200 words"). PromptGen bakes these into the system prompt rather than relying on you to remember.

Pick model target + output format

Two dropdowns control the shape of the prompt:

Field	When to pick which
Model target	Claude for Anthropic models · GPT for OpenAI · Gemini for Google · Generic if you'll port the prompt around.
Output format	JSON when you'll parse the output programmatically · Markdown for human-readable structured output · Plain for free-form prose.

Why this matters: Claude responds best to XML tags + explicit role framing. GPT prefers numbered constraints. Gemini wants short, declarative sentences. PromptGen adjusts wording, not just labels.

Generate (≈10 seconds, 2 credits)

Hit Generate. You get three blocks:

System prompt — the persona, rules, output format, and constraints. Paste this into the system field of your LLM call.
User prompt template — the user-turn template with {{placeholders}} for the actual input.
Usage guidance — model temperature, max-tokens, common gotchas, and one or two example invocations.

Export and ship

Hit Export as Markdown to download the whole prompt pack as promptgen-output.md. This file is paste-ready for:

Your code editor (VS Code / Cursor) — drop into prompts/ directory
Claude Projects / GPT Custom Instructions / Gemini Gem builder
n8n / Zapier / Make automations — paste into the AI node

Each generate is also logged to /portal/ → My Apps so you can re-find a prompt you wrote last week without regenerating.

What this app is not for

Brand-voice copy — use BrandVoice instead.
Marketing taglines — use TaglineGen.
Chatbot personas — use OperatorChat; it adds 5 sample exchanges + voice notes that PromptGen doesn't.

What's next? The prompts you generate need to live somewhere. If you're shipping AI features in a product, the "Usage guidance" block tells you which temperature + max_tokens to set. If you're using Claude/GPT directly through chat, paste the system prompt as your custom instructions and the user template as your message — replace placeholders inline.