Custom agent

What is Custom agent?

The Custom agent provides a flexible framework to create and configure your own AI agents tailored to specific needs. It allows you to define the agent's behavior, knowledge sources, and even its appearance, giving you significant control over its functionality.

The Custom agent offers the following extensive customization options:

• Customizable knowledge base: You can select multiple workspaces and even websites as knowledge sources for your agent, allowing it to draw information from diverse locations.

• Tailored responses: You can define a system prompt to guide the agent's behavior and set specific instructions for how it should format its responses (e.g., using bullet points, specific tone).

• Conversation starters: You can pre-define conversation starters that appear as clickable prompts, helping users initiate interactions with your Custom agent.

• Branding and appearance: You can upload a logo for your Custom agent, and if published to your website, its appearance can be further customized using CSS.

• Integration with external websites: Custom agents can be published and integrated into your own company's website using provided script tags, allowing you to host an AI assistant tailored to your specific content.

Configuring and creating the Custom agent

The process of configuring a Custom agent involves both creation and detailed setup:

1. Navigate to Agents section and click Explore.

   

2. In the Explore page, click Create agent. The Create Agent page is displayed.

3. Enter the following details:

   • Basic Info:

     

     • Name and Description: Give your agent a clear identity with a unique name and description.

     • Upload Logo: Upload your company logo. This unlocks additional customizability, allowing you to set specific background and text colors so the interface feels like an extension of your brand.

   • Deep Research: Toggle the Deep Research option if your agent needs to perform complex, multi-step analysis or browse the web for real-time information.

   • Advance Settings: Configure the following:

     
     • Enable and configure which LLM models users can select for this agent.
     • Enable voice features like Speech-to-Text or Text-to-Speech if you want to allow your agent to process speech input and generate speech output.
     • Provide the place holder text to be shown in the chat input box.

   • System Prompting: Define the agent’s personality. This is where you instruct it on how to speak and how to handle specific client inquiries.

     

   • Tools & Capabilities: Configure MCP servers and other tools for your agent.

     
     • To make your agent useful, you must give it a source of truth. Scroll to the Workspaces section. Select the workspace containing your documents. The agent will now be able to search these documents and provide citations (links to the exact source) in its responses.
     • Select the required Connectors. You can also click the New Connector button and create new credentials for the reviewed and recommended tools.
     • Add Model Context Protocol (MCP)servers that provide additional capabilities to your agent.
     • Turn on the Web Search button to enable web search and browser capabilities for your agent to access live internet information.

Using the Custom agent

Once your Custom agent is configured and saved or published, users can interact with it based on its defined knowledge base and instructions.

• Within the platform: Select your Custom agent from the Explore section. You can then ask questions, and the agent will respond based on the content of its configured workspaces and websites, adhering to the system prompt and instructions you've set.

• On your website: If you have published the agent and embedded its script, it will appear as an interactive element (e.g., a chat popup) on your website. Users can then ask questions related to your website's content or the specific knowledge you have provided.