Discover Dust
Discover Dust
20 min
Gaëlle CaplierYou are now ready to take your first steps on Dust. Let's start with a quick product tour!
1. The Dust homepage
Go to
app.dust.tt to access the Dust homepage. It is organized in three main sections1. The input bar, where you can chat with your agents
2. Your history of previous conversations
3. Your agent library, with all available agents on your workspace
2. Your team of agents
In Dust, there are two types of agents: global agents, and custom agents. Global agents are multipurpose and available by default on your workspace, while custom agents are specialized and are built by yourself or your colleagues.
2.1. Global agents, your multipurpose assistants
Global agents are available by default in Dust, you don't need to build them yourself.
2.1.1. Out-of-the-box AI models
In Dust, you can directly access and interact with all of the leading AI models (LLMs) on the market. We update the list of available models in Dust after each new major release, so you always have access to state-of-the-art models.
Those models are
@gpt, @claude, @mistral, @gemini, and more. These are like using ChatGPT, Claude, or other AI tools directly: powerful, but without any knowledge of your specific context.⚠️ If some of those models are not available for you, that’s because your administrators have made the decision to deactivate them for your workspace.
2.1.2. @help, your Dust helper
@help is here to... help you! Ask it anything about how to use Dust, and it will find the answer for you in the Dust documentation.2.1.3. @dust, your everyday copilot
The
@dust agent is connected to all data sources in your workspace, and it learns more about you with each of your interactions. You can use it for any kind of task that does not require advanced specialization - company research, email writing, etc.2.1.4. @deep-dive, your in-depth research agent
@deep-dive is made for long-running research on complex topics. It is really 5+ agents in a trenchcoat: when you ask a complex question to @deep-dive, it will decompose it into smaller sub-tasks and spin up sub-agents, each focusing on one of those sub-tasks.@deep-dive has access to both your company knowledge and to the internet. It is excellent at tasks like competitive intelligence and market research.⚠️ For very complex queries, it can take 30+ minutes to finish researching and provide an answer.
2.2. Custom agents
Custom agents are built by yourself or your colleagues. Think of custom agents as LLMs that have been given a job description, trained on your company context, and equipped with the right tools.
Why should you build custom agents?
LLMs - like GPT, claude, gemini, and the others - are powerful because they have been trained on massive amounts of data. They are occasionally described as “the average of all human knowledge”; but in your professional usage, you need more than average: you need specialized context-aware AI. That’s where Dust custom agents come in.
Custom agents leverage the usual state-of-the art LLMs, but with added context and capabilities. They have access to your company data and can take actions for you in your usual work applications. Specifically, they have:
- Custom instructions - Specific roles and processes tailored to your needs
- Custom knowledge sources - Access to relevant company data
- Custom capabilities - Ability to search web, create visualizations, draft emails, etc.
3. Interacting with AI in Dust
3.1. Your modes of interaction
- Calling an agent: Type
@followed by the agent name.- Ex:
@claude-4.5-sonnet what is machine learning?
- Parallelizing agents: by mentioning more than one agent in a single message, you are sending your query in parallel to all of them. You can then compare their answers.
- Ex:
@gpt5 @claude-4.5-sonnet @mistral explain quantum computing
- Chaining agents: you can chain different agents in a single conversation. All agents will be aware of the context of the ongoing conversation, allowing you to chain different agents to execute a multi-step process.
- Ex: you can ask
@gpt5to synthetize and expand on answers provided by other LLMs - Ex: you can use a custom agent
@meetingSummaryto generate a meeting recap based on your notes, then call another custom agent@meetingFollowUpto draft an email follow-up to your attendees with this recap.
- Message edition: Made a typo? You can edit or delete messages after sending.
- Mentions: use @ to add a colleague to a conversation or notify them in shared conversations.
3.2. Expand your agents' capabilities in-conversation
You are not limited by your agents' out-of-the-box capabilities: you can extend them directly in the context of your conversations, by adding more knowledge and more capabilities directly from the input bar.
⚠️ Any capability or knowledge that is added to an agent in a conversation will only be available during this conversation: you have not permanently updated the agent with the new capabilities, meaning that if you start a new conversation from scratch with the same agent, you will have to add those capabilities again.
4. Finding the right agent
There are several ways to find the right agent for your use case:
- Browse by tags: Agents can be organized by function, department, or use case
- Star your favorites: Add frequently-used agents to favorites for quick access
- Search bar: Type keywords to filter agents by name or description
- Ask @help: Not sure what exists? Ask "@help what agents are available for [your use case]?"
If you still can’t find an agent to solve your issue, then it may be time to build a custom agent yourself!
Test Your Knowledge
Answer 5 questions — score 3 or more to pass
Ready to test your understanding of this chapter?