> ## Documentation Index
> Fetch the complete documentation index at: https://docs.monk.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Talking to Monk

> How to phrase requests, what Monk asks you, and where you approve

You talk to Monk from your coding agent. Start a message with `/monk` and say what you want in plain words. There are no flags to learn.

## Before you start

* Monk is installed in your coding agent. See [Install](/getting-started/get-started).
* You are signed in to Monk from the agent. Signing in on the website is a separate step.

## Ask Monk

Start with a plan. It is the safest first prompt for any project:

```
/monk plan a deployment for this app
```

Then ask for what you need, in your own words:

```
/monk deploy this to AWS in eu-west-1
/monk show me the logs for the api
/monk what is this cluster costing this month?
/monk set up CI/CD for this repo
```

Vague requests work. Monk asks for what it is missing, such as a cloud or a region. More detail up front means fewer questions.

## How work gets done

1. **Your coding agent** reads the code and writes the Dockerfiles and Monk's `MANIFEST`.
2. **Monk** validates them, builds a plan, and shows it to you.
3. **You** review and approve the plan in the local dashboard.
4. **Monk** provisions, deploys and keeps a model of the running app, so later requests start from what is actually running.

Monk does not edit your application code. If a deploy needs a code change, your coding agent makes it in the usual way.

## What you review

Anything that creates, changes or deletes infrastructure, pushes secrets or runs an action on a workload opens an approval in the [local dashboard](/getting-started/local-dashboard) at `http://127.0.0.1:7419`. The approval shows what will change and, for new infrastructure, the estimated cost. Nothing happens until you approve.

## Credentials never go in chat

When Monk needs a cloud key, an API token or a password, it opens a form in the local dashboard. Type or upload the value there. Don't paste credentials into the conversation.

Values are stored in your OS keychain (or an encrypted file where there is no keychain) and pushed to your cluster only when a deploy needs them.

## Answering Monk's questions

Answer directly ("AWS"), ask for a recommendation ("what would you pick for a small API?"), or give context ("this is a staging copy, keep it small"). Monk's suggestions are suggestions. The plan you approve is what runs.

## Reporting a problem

```
/monk report a bug
```

This sends a report with details about your environment. There is no reply channel, so include what you expected and what happened.

## Next

* [Deploy an existing app](/guides/deploy-existing-app)
* [Start a new app](/guides/start-new-app)
* [How Monk works](/concepts/how-monk-works)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.