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

# Chat Interface

> How to use Brunel's chat interface to build high-quality plans with AI.

The chat interface is where you collaborate with Brunel's AI to build structured plans. It's designed to feel familiar — a message thread with persistent history — but purpose-built for planning work rather than general conversation.

***

## The Interface

The workspace is organized into three areas:

* **Sidebar** — your projects and sessions. Click any session to open it as a tab
* **Session tabs** — multiple sessions open simultaneously at the top of the workspace. Switch between them instantly
* **Chat area** — the active session's message thread, input box, and file panel

***

## Writing a Good First Message

The quality of your plan depends heavily on your first message. The more context you give upfront, the less back-and-forth is needed and the more useful the exported plan will be for your coding agent.

A strong first message includes:

* **What you're building or fixing** — a clear description of the task
* **Relevant constraints** — existing services, patterns, or libraries that must be used
* **What to avoid** — approaches that won't work for your codebase or team
* **Acceptance criteria** — how you'll know the work is done

<CodeGroup>
  ```text Weak first message theme={null}
  Add rate limiting to our API
  ```

  ```text Strong first message theme={null}
  Add rate limiting to our API endpoints. We have an existing
  RateLimiterService backed by Redis that must be used — don't
  create a new one. Free tier users are limited to 100 req/hour,
  paid users to 1000/hour. Limits need to be configurable per
  endpoint. Must integrate with our existing UserAuthMiddleware
  to check tier. Acceptance: tests for both tiers pass, config
  is per-endpoint.
  ```
</CodeGroup>

***

## Quick Reply Buttons

As the conversation progresses, Brunel surfaces **Quick Reply** buttons — pre-built prompts that help you move the planning conversation forward without typing. These might include options like "Add more detail on dependencies", "Show acceptance criteria", or "I'm ready to export this plan." Click any button to send that message instantly.

***

## Tool Call Indicators and Thinking Blocks

When Brunel's AI is doing background work — reading context files, searching for patterns, or processing complex requirements — you'll see a **tool call indicator** showing what it's working on. This is normal and means the AI is gathering the information it needs before responding.

For complex planning tasks, you may also see a **thinking block** — a collapsible section showing the AI's reasoning process before it delivers its response. You can expand this to understand how the AI arrived at its plan.

***

## Message Actions

Every message in the chat supports:

* **Copy** — click the copy icon on any message to copy its full content to your clipboard. Useful for pasting plan text directly into a coding agent
* **Timestamps** — hover over any message to see its exact timestamp. Messages display relative time by default (e.g., "5m ago")

***

## Uploading Files in Chat

You can attach files directly in the message input using the **attachment button** (paperclip icon). Files uploaded this way are added to the session's Context Files and immediately available to the AI in subsequent messages.

For managing multiple context files, use the dedicated **Context Files panel** — see [Context Files](/planning/context-files).

***

## Planning Complete Banner

When the AI has finished building a plan, a green **Planning Complete** banner appears above the input area. It shows:

* A **confidence score** indicating how complete the plan is
* A brief **summary** of what the plan covers
* A **Review Docs** button — opens the planning documents folder on your machine (appears when docs have been written to disk)
* A **Proceed to Implementation** button — moves the session to the Execution phase

***

## Sending Messages While the Agent is Working

You can type and send a message while the agent is still responding. The message will be **queued** and delivered automatically once the current response completes. A queued message indicator appears above the input area showing how many messages are waiting. Hover over it to preview the queued content.

***

## Sub-Agent Activity

For complex tasks, the planning agent may spawn sub-agents to handle work in parallel — for example, a research sub-agent or a verification sub-agent. When a sub-agent is active, you'll see an indicator in the session. Click it to open the **Sub-Agent Panel** on the right side of the screen, which shows real-time streaming output from the sub-agent, including tool calls and text output.

The panel closes automatically when you switch sessions.

***

## Context Window Usage

At the bottom of the session, a usage indicator shows how much of the agent's context window is currently in use. This is useful for long planning sessions — if the context fills up, the system will compact conversation history automatically to keep the session running smoothly.