> ## Documentation Index
> Fetch the complete documentation index at: https://docs.replit.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Prompting
> Give Replit clear direction in a chat and refine the result with useful follow-up requests.
export const PromptingCover = () =>
;
};
Prompting means telling Replit the result you need in plain language. You don't need a perfect prompt. A short question or a rough idea is often enough to begin, and you can develop the brief through follow-ups.
Use prompts to ask questions, research topics, plan work, schedule Routines, write documents, or create apps, slides, and designs.
Here are three simple steps to prompt effectively:
1. Start simple.
2. Share useful context.
3. Ask follow-up questions and refine the chat.
Let's look at each one.
## Start simple
A short request can establish the direction:
What should a small team include in a weekly project update?
Add context, boundaries, and an intended use when the task needs them:
Research how small product teams run weekly planning. Use current web sources
and the attached team notes. Compare three approaches, keep the summary under
500 words, and end with a recommendation I can share at Monday's planning
meeting.
The richer prompt is not automatically better. The extra details are useful because they change the result.
## Give Replit useful direction
When a task needs more direction, focus on what matters. Include only the details Replit needs:
* **State the outcome.** Ask for an answer, analysis, plan, Routine, or something you want to create.
* **Add useful context.** Point Replit to relevant files, [integrations](/features/integrations/overview), web sources, examples, feedback, images, or screenshots.
* **State important boundaries.** Say what Replit should preserve, prioritize, or leave out. Prefer direct instructions that describe what you want.
* **Explain how you will use the result.** Name the audience, decision, format, or next action when it affects the work.
These are options, not a required formula. Start with the details you know, then add more when they help.
## Refine in the chat
Use follow-up requests to add context, respond to what Replit produced, or move the work toward an outcome:
* “Show the evidence behind the second finding.”
* “Use the error message I added and revise the likely cause.”
* “Keep the analysis, but turn it into five slides for leadership.”
* “Create a simple tracker from the recommended next steps.”
* “Explore three design directions before I choose one to build.”
* “Run this review every Monday as a Routine.”
Working incrementally helps you review each result before deciding what Replit should do next.
### Keep going from one result to the next
A follow-up can turn one result into the next useful output. Replit keeps the context from earlier in the chat, so you don't need to repeat the brief:
| Start with | Then keep going |
| - | - |
| “Help me compare these three pricing options.” | “Build a calculator so I can test different scenarios.” |
| “What keeps coming up in these survey responses?” | “Build a prototype that addresses the biggest problem.” |
| “Help me shape this idea for a weekend running club.” | “Build a signup page I can publish.” |
| “Explain how compound interest works.” | “Make an interactive example where I can change the numbers.” |
| “What should I ask customers before preparing a quote?” | “Build an intake app to collect these details.” |
For more guidance, explore the [prompt box](/home/prompt-box), learn about [context management](/learn/foundations/context-management), or [see detailed prompting examples](/learn/effective-prompting).
## Next steps
Continue from your prompt with one of these paths:
Learn how to [steer or queue follow-up messages](/features/agent/steer-and-queue-messages) while Replit works.
Once the request is clear, learn how to [add context](/chat/add-context) so Replit has the relevant information and tools for the task.