Creating and managing Knowledge Agents

👥
Access Required
Knowledge Agents are available on Guru’s All-in-One or Enterprise plans.
Only **Admins, Creators, and relevant custom roles can create Knowledge Agents. A Knowledge Agent Owner or relevantcustom role can be assigned to manage settings and permissions.
Before creating a Knowledge Agent, ensure you have connected Sources or created content in Guru.

Not sure if you need a Knowledge Agent? Ask yourself and your team the following:

Are your search results and suggested answers cluttered with less relevant content?
Do you want conversational AI interactions that understand context and can handle follow-up questions naturally?
Are multiple teams using Guru with unique needs and preferred sources?
Do you want Guru to generate specific, tailored outputs (e.g., customer response templates or curated prep lists) to make your team more efficient?
Are there repetitive questions in Slack channels that Guru could answer automatically through chat?

If you answered "yes" to any of these, a Knowledge Agent can help.

Check out the video below for more:

Top 5 Reasons to Create a Knowledge Agent

1. Enable Natural, Conversational AI

Knowledge Agents power chat conversations that understand context, remember what you've discussed, and provide follow-up responses that build on previous questions — just like talking to a knowledgeable teammate.

2. Drive Internal Efficiency

Knowledge Agents surface precise, context-aware answers faster in both search and chat — minimizing time spent searching and increasing team productivity through natural interactions. With Research Mode enabled, agents can perform comprehensive research combining internal knowledge with real-time web search for thorough, well-cited responses.

3. Reduce Expert Burnout

By handling repeat questions and providing accurate responses through conversational AI, agents free up your subject matter experts for higher-impact work.

4. Customize by Team

Each team can create agents tailored to their roles and tools — ensuring answers and actions align with how they work, whether in search or chat conversations.

Example: A Support team agent might pull only from the Help Center and suggest ticket macros during chat, while a Sales agent might prioritize battlecards and provide conversational call prep assistance.

5. Improve Relevance Across All Interactions

Agents can be configured to use only the most relevant sources and templates — refining both manual searches and AI-powered chat conversations.

Example: A Support team agent might pull only from the Help Center and suggest ticket macros during chat, while a Sales agent might prioritize battlecards and provide conversational call prep assistance.

Building a Knowledge Agent

Check out the video below on building a Knowledge Agent:

In Guru's web app, go to Manage > Knowledge Agents.
Click Add Knowledge Agent.
Enter a name and upload an image for your Agent.
Add a clear description explaining the agent's purpose and the types of questions it's designed to answer. This helps users understand when to use this agent and what to expect from conversations.

💡
Tip
Use clear, descriptive summaries to help users quickly find and choose the right Knowledge Agent for their needs. When managing multiple agents, good descriptions reduce confusion and improve adoption.

Choose to sync from:
- All Sources – includes current and future sources.
- Specific Sources – click Next: Select Sources, then select the relevant ones and proceed to Next: Set Permissions.
Assign Viewers and relevant custom roles (users or Groups).
Assign a Knowledge Agent Owner.
Click Create Agent.

Enabling Research Mode for comprehensive responses

Research Mode allows your Knowledge Agent to perform deeper, multi-step research, combining internal knowledge with real-time web search for comprehensive, well-cited answers in both search and chat conversations.

Turn on Research Mode

Go to Manage > Knowledge Agents and select your agent.
Navigate to the Behavior tab.
Toggle Perform research to enable Research Mode.
Click Save.

💡
Tip
Enable Research Mode for Knowledge Agents that support complex queries, need up-to-date external information, or when users need thorough insights rather than quick answers.

With Research Mode enabled, your Knowledge Agent will:

Combine multiple sources for more complete answers
Access current information through web search when appropriate
Provide citations and sources for research-backed responses
Handle complex, multi-part questions more effectively in chat conversations

Understanding Guru's Default Knowledge Agent

Every Guru workspace includes a default Knowledge Agent named "Guru" that appears in the search bar and provides AI-generated answers in both search and chat. This agent serves as the foundation for your team's AI-powered knowledge experience.

By default, this agent is connected to all sources and permissions based on the access to those sources. In other words, users will only see what they have access to, whether in search results or chat conversations.

Admin control over the default agent:

Admins have full control over the default Knowledge Agent and can customize it to match their organization's needs:

Renaming and rebranding the default agent:

Go to Manage > Knowledge Agents
Locate the default "Guru" agent
Click to open the agent settings
Update the Name field with your preferred name
Add or update the description to explain its purpose and scope
Upload a custom Image to rebrand the agent
Click Save

Controlling which sources the default agent uses:

By default, the agent uses "All Sources," but Admins can restrict this to specific sources:
Go to Manage > Knowledge Agents
Select the default Knowledge Agent
Navigate to the Sources tab
Choose between:
- All Sources – includes current and future sources (default setting)
- Specific Sources – select only the sources you want this agent to use
If selecting specific sources, choose from your connected Collections and Sources
Click Save

💡
Tip
Customizing your default agent's name, description, and sources can help create a more branded experience and ensure users get the most relevant results for their primary workflows, whether they're searching or chatting.

Managing Sources for a Knowledge Agent

From the Sources tab, modify which sources the Agent uses for both search and chat responses.
Users need access to both the Knowledge Agent and its linked Sources or Collections to receive answers in any format.
In the Files tab, you can upload individual .PDF or .TXT files as Sources.

✍️
Note
If you are connecting a private Slack channel as a Source, you must be a Slack Admin.

Setting permissions

Manage access from the Permissions tab:

Add Groups or users as Viewers.
Promote Viewers to a relevant custom role.
Assign a Knowledge Agent Owner.

✍️
Note
Admins, Knowledge Agent Owners, and relevant custom roles do not automatically have access to all Sources. They must be permissioned to receive results and see sources in both search and chat experiences.

Roles and permissions

Viewers can:

Use the Agent in Guru search, chat, and Slack.
View assigned Agents on the results page.
Have natural conversations with Knowledge Agents.
See their own questions in the AI Agent Center.

Admins, Knowledge Agent Owners, and relevant custom roles can:

Manage prompts, permissions, and Slack/Page integrations.
Enable or disable Research Mode.
Delete or update Knowledge Agents.
Create custom prompts for chat and search responses.

Creating a custom prompt

Prompts tell your Knowledge Agent how to answer questions in a specific tone or format, whether in search results or chat conversations.

From the Prompt tab, customize and test your Agent's prompt.
You can revert to the default prompt at any time.

Changing the output message when no answer is found

Within the same Prompt menu, Admins can write their own fallback message in markdown, which gives full control over what's shown to end users when no answer is found in either search or chat. This means you can format the message with headers, links, images, and more—just like a Card. It's a simple way to guide users, set expectations, or point them to helpful resources.

Go to Manage > Knowledge Agents and locate the Agent you want to adjust
Navigate to the Prompts tab in that Knowledge Agent
Click Edit and adjust the No answer found message as needed
Click Test and enter a question you know will not generate an Answer to see your custom response generated
Click Save when ready

Adding a Knowledge Agent to your workflow

Adding to Slack channels for chat conversations

From the Slack tab:

Enable the Agent in Slack.
Assign the Agent to specific Guru-connected channels for natural chat conversations.
Toggle Limit responses to sources shared with All Members to ensure sensitive content isn't shared in public spaces during chat interactions.

💡
Tip
Seeing different answers in Slack vs. Guru? If responses vary between search and chat, it may be because Limit responses to sources shared with All Members is enabled in Slack. To update, go to Manage > Knowledge Agents.

Adding to Guru Pages

You can assign Knowledge Agents to specific Pages to power both search and chat experiences:

Open the Page and click the Edit icon.
Set the default Knowledge Agent.

Adding to the Guru Extension when over a specific webpage

You can assign a Knowledge Agent to appear as the default in the Guru Extension when visiting a specific webpage, enabling contextual chat and search.

Click Website > Add a Website
Fill in the display name and URL
Click "Add Website"

Using the API

Admins and relevant custom roles can copy the search and answer endpoints from the Agent to use with Guru's API for programmatic access to both search and chat capabilities. View API documentation for implementation details.

✍️
Note
Don't forget to toggle on Activate this Knowledge Agent once it's ready to use in search and chat.

Frequently asked questions

What's the difference between Knowledge Agents in search vs. chat? Knowledge Agents power both experiences but excel in chat by enabling natural, contextual conversations with follow-up questions. Search provides quick answers, while chat allows for deeper exploration of topics.

Why should I add a description to my Knowledge Agent? Clear descriptions help users understand the agent's purpose and select the right one for their questions, especially when multiple agents are available. This reduces confusion and improves user adoption.

Can Knowledge Agents remember previous parts of our conversation? Yes! In chat conversations, Knowledge Agents maintain context and can reference earlier parts of your discussion to provide more relevant follow-up responses.

Can I assign more than one Knowledge Agent to a Slack channel? No. Each Slack channel supports one Knowledge Agent at a time. You can assign the same Agent to multiple channels.

If a Group is given access to a Knowledge Agent, but that Group is not given access to the Sources connected to that Knowledge Agent, will the Knowledge Agent pull responses from those Sources when asked?

No. The Collection or Source will be filtered out of the results if a user does not have access to that Collection/Source.