Creating Agents – Here’s How

(the content of this page is available in German and English)

🤖What are Agents?

Agents are AI-powered tools that can take on tasks for you. For example, agents can help you summarize information, make decisions based on that information, handle recurring tasks, and more. They assist you in making your processes even more efficient.

🧠AI Dialogue and Agents – When to Use What?

You are already familiar with AI dialogue: Quickly asking a question, gathering ideas, or getting technical support – it’s simple and flexible. Have you often wished you could save certain templates? That’s exactly what agents are for! Agents work similarly to AI dialogue but go a step further. You can provide and save them with fixed instructions and knowledge sources. This way, you start directly with the appropriate context without having to formulate a new prompt each time. In short: AI dialogue is perfect for spontaneous questions and creative impulses – both with your company knowledge and for general topics. Agents, on the other hand, are your automated helpers that take over recurring tasks and work exactly according to your specifications.

🔍 Where to Find Agents?

It’s simple: Click on Agents in the sidebar. Here you can:

Discover Agents

• Try out Shared Agents

• Mark Favourites

• Share Agents

Use the search bar or filters to find agents. The sections “Favorites,” “My Agents,” and “Shared Agents” make it easy to keep track.

💡 Tip: Click the ✏️ icon on the agent card to configure agents, or the 🗑️ to delete (only possible for your own agents). You can share agents with others using the paper airplane.

⭐ Creating Agents – Things Are Getting Exciting!

Want to create your first agent? It’s easy. Click “Create” in the “AI Agents” menu, and you’re off! 💡 Tip: Need a bit of inspiration on what an agent could do for you? Here’s the agent library with examples, prompts, etc. Just browse, copy, and get started 📚✨

👤 Name

What should your agent be called? You’ll see the name later in the overview.

Description

Here you can provide a short description of the agent’s purpose. The description does not influence the agent’s behavior but will be displayed in the agent overview.

Instructions

This is the important part – what should your agent do?

Instructions determine how your agent will behave. The clearer you describe what your agent should take on, the better it will do its job. In the instructions, you can specify whether your agent should follow certain rules, what tone it should use for responses, etc.

Got more to say? No worries, we’ve got that covered. Simply use the lower right corner to drag the input box bigger

Example Prompts

Example prompts are questions that could be asked of your agent. These prompts are displayed directly when the agent is opened and are clickable so the agent can get started immediately.

That neat little trick from the Instructions that allows you to drag the box bigger? Works here as well!

Deep Research Mode

Your agent should do more than just give quick answers – it should really dig deep? Then the Deep Research Mode is exactly what you need!

With Deep Research Mode, your agent can perform multi-step, complex research based on your selected data sources. It systematically searches through your internal knowledge and presents the results in a structured summary – perfect for comprehensive analyses, reports, or questions that require multiple sources.

Simply toggle the switch next to "Deep Research Mode" to activate it.

Works with Web Search: Deep Research Mode can also be combined with web search. This way, your agent doesn't just search internal data sources – it can also pull in external information from the web as part of its research. This is especially useful when your internal knowledge alone isn't enough and you need current market data, industry insights, or publicly available sources to complement your findings.

Good to know: When Deep Research Mode is activated, you cannot select an AI model (LLM) and image generation is not available. Your agent will focus entirely on research.

When is Deep Research Mode worth it?

• You need a comprehensive summary across multiple documents

• You want to answer a complex question that requires various sources

• You want to create a detailed report based on your company knowledge

• You want to combine internal insights with external information from the web – e.g., compare internal product data with current market trends

💡 Tip: Enable web search alongside Deep Research Mode when your agent should compare internal processes with current best practices or regulatory requirements, for example.

AI Image Generation

Your agent doesn't just write text – it can also generate images! With AI Image Generation, your agent can create visuals on demand, such as illustrations, diagrams, or creative drafts.

Simply toggle the switch next to "AI Image Generation" to activate it.

The model used for image generation is set by your admin – so you don't need to worry about a thing. Just enable it and get started!

What is image generation great for?

• Social media posts: Your agent creates text and a matching image in one step

• Presentations: Quickly generate an illustration or a visual diagram

• Creative projects: Visualize ideas, create moodboards, or sketch design drafts

• Marketing materials: Generate visuals for newsletters, blog posts, or campaign concepts

💡 Tip: Combine image generation with clear instructions in your agent – e.g., "For every social media post idea, also create a matching image suggestion." This turns your agent into a creative all-rounder!

⚠️ Good to know: Image generation is not available when Deep Research Mode is activated.

Configuring the Knowledge Base

Here is where your agent gets really smart. In the Knowledge Base, you have the option to add company knowledge, your own documents, or web pages, allowing the agent to work even more precisely. It is currently not possible to combine all knowledge sources, meaning you cannot activate web search and internal knowledge at the same time. Just click the toggle to use this option.

The Knowledge Base

To use internal knowledge, simply click the toggle, and you’re good to go. With “Configure Knowledge Database,” you can browse folders and data structures and make them available to your agent directly.

The Knowledge Base

To use internal knowledge, simply click the toggle, and you’re good to go. With “Configure Knowledge Database,” you can browse folders and data structures and make them available to your agent directly.

Once you’re inside the Data Sources menu, you can either use the Search Bar to view through different Data Sources by name, or use the Data Sources dropdown menu to select a specific source. You can mix across different Data Sources - be it emails, a specific Sharepoint folder, something in your Jira - we got you covered.

Use the toggle “Folder Only” to view and search for for folders. By clicking the box next to a folder or Data Source name, the entire content will be moved to the right and you will see it under Knowledge Base Preview. Success! Your knowledge soure is now ready to make your agent even smarter. And if you want to delete something again, it’s just as easy. Simply click on the little 🗑️and your agent will be non the wiser.

Note that it’s not possible to search for specific files in the Data Sources view. If you wish to attach a specific document to your agents, please use the Uploading Files menu.

Uploading Files

Under “Uploading Files,” you can select files that the agent should use. Just activate the option with the toggle and select one or more files.

Web Search

The web search allows your agent to access knowledge from the internet in real-time. Simply provide one or more URLs (the address from the browser) to add web pages. You can add a maximum of five websites.

Selecting Language Model (LLM)

Here you can choose which language model the agent accesses. This selection is optional and not available to all customers. Here are some examples.

1. GPT 5

   Strengths: Highly versatile, strong reasoning and contextual understanding, excellent at professional knowledge work tasks
   Areas of Application: Text generation, chatbots, summaries, analysis, complex multi-step tasks

2.   Claude Opus 4.6

   Strengths: State-of-the-art coding performance, hybrid reasoning with extended thinking, exceptional at complex multi-step analysis and agentic workflows
   Areas of Application: Coding, enterprise workflows, complex research, creative writing, autonomous agents

3. Gemini 3.1 Pro

   Strengths: Multimodal (text, image, video, audio), very large context window, strong integration capabilities
   Areas of Application: Multimodal analysis, long document processing, data-heavy tasks, visual understanding

   

Visibility – Who Can See Your Agent?

By default, your agents are only visible to you – you can tell by the small lock symbol 🔒. If you are an admin, you can share your agent with the entire organization: Just click on “Shared with my Organization” – and it’s available to everyone. It’s just as easy to make it private again: With one click on “Visible only for me” 🕵️.

From Idea to Agent – Just Click

Finally, don’t forget: Click “Save” – and your agent comes to life! 🤖✨

And don’t worry: If you want to make changes later, you can open your agent anytime via “Edit” and give it the final touches.