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

# What is The Smith?

> Guild's built-in support agent and chat landing experience.

Smith is Guild's built-in support agent. It replaces the legacy dashboard as the default landing experience when you open a workspace, putting chat front and center so you can start working immediately.

## What Smith does

Smith is always available in every workspace. You can ask it to:

* Answer questions about Guild features, agents, and configuration
* Help you build, debug, and publish agents
* Walk you through setting up integrations and triggers
* Troubleshoot errors in your sessions or agent code

## Starting a conversation

When you open a workspace, Smith's chat interface is the first thing you see. Type a message and press **Enter** to begin.

You can also open Smith at any time from the left sidebar by clicking the **Smith** icon (the chat bubble at the top of the nav).

## Switching agents

Smith is the default agent, but it is not the only one. Use **@mention** to route a message to a different agent without leaving the chat interface. See [Agent mentions](/platform/sessions#agent-mentions).

## Workspace context

Smith has read access to your workspace configuration — installed agents, triggers, integrations, and recent sessions — so it can give context-aware answers without you having to copy-paste configuration details.

## Safety limits

Smith includes internal protections that stop a single conversation from looping or running up cost:

* **Turn budget.** Smith runs at most 20 LLM turns per user message. When a request hits this limit, Smith stops, saves its progress, and tells you it will not keep retrying automatically.
* **Duplicate-response stop.** If Smith produces the same final answer twice, it stops instead of repeating the attempt.
* **Unsupported integration stop.** When you ask Smith to set up credentials for a service it cannot resolve to a supported integration, it stops after one discovery pass, names the unsupported service as the blocker, and does not retry.

These limits apply on top of the platform-wide [execution limits](/reference/limits).