Skip to main content
Prompt Generator is a two-pane workspace:
  • Left: describe your goal (“idea”) and choose a target model
  • Right: iterate in chat, then save the version you want to your Library

​
Core concepts

  • Target model: the model the prompt is optimized for (e.g. GPT, Claude, Gemini).
  • Chat model: the assistant you use to iterate on the prompt inside the chat.
You can optimize for one model and iterate with another.
Each time you generate, Prompt Builder creates a new chat thread so you can revisit iterations later.
Use Save to store the exact version you want in My Prompts, so you can reuse it later (and run it in Assistant).

​
Generate a prompt

1

Describe the outcome

Write what you want to accomplish. Include:
  • Audience / role
  • Context / constraints
  • Desired output format
If you know the output format, state it explicitly: “Return JSON with keys …” or “Return a table with columns …”.
2

Choose a target model

Pick the model you want the prompt optimized for.
3

Click Generate

Prompt Builder creates a new chat thread and posts the first prompt version into the chat.

​
Refine in chat

Use the chat on the right to iterate quickly. Good refinement prompts include:
  • “Make it shorter and more direct”
  • “Add step-by-step instructions and acceptance criteria”
  • “Ask clarifying questions before answering”
  • “Add 3 examples and 3 edge cases”
  • “Rewrite for a different tone: friendly, formal, terse”
When you like a version, click Save to add it to your Library.

​
Reuse saved prompts

You can insert an existing prompt from My Prompts into the composer and keep iterating (no need to start from a fresh idea).

​
Next steps

Prompt Optimizer

Improve an existing prompt and keep a history of versions.

Prompt Library

Organize saved prompts and run them in Assistant.

Prompt Assistant

Chat-first help for drafting, adapting, and refining prompts.