Creating a successful agent in elvex requires more than just filling out a form—each setting plays a critical role in how your agent performs, how users interact with it, and whether it delivers accurate, helpful responses. This guide explains each agent setting and how it impacts your agent's success.

Understanding how to configure these settings properly will help you create agents that are reliable, accurate, and tailored to your specific use cases.

There are three ways to create an agent:

Check out example agents as well as our template documentation.

The Overview section establishes your agent's identity and helps users discover and understand its purpose.

What it is: A short, descriptive name displayed on the agent's page

Why it matters: A clear, specific name helps users quickly identify the right agent for their task. Generic names like "Helper" or "Agent 1" make it difficult for users to choose between multiple agents.

Best practice: Use names that describe the agent's function or domain, such as "Sales Proposal Writer," "HR Policy Expert," or "Financial Data Analyst."

What it is: Keywords that associate your agent with specific categories

Why it matters: Tags enable users to filter and discover agents relevant to their needs. In organizations with dozens of agents, tags are essential for findability.

Best practice: Use consistent, meaningful tags that reflect the agent's department, function, or use case. Company Admins can manage available tags to maintain consistency.

What it is: An explanation of the agent's purpose, goals, and capabilities that serves as the agent's "north star"

Why it matters: The description serves two critical purposes:

A well-written description keeps your agent focused on its intended purpose and helps it make better decisions about how to respond to user requests.

Best practice: Write 2-3 sentences that answer:

Think of this as the agent's mission statement—it should be clear enough to guide both users and the AI model itself.

Example: "This agent helps sales teams create customized proposals by pulling from our product catalog and pricing database. It's designed for account executives who need to quickly generate professional proposals for enterprise clients."

What it is: Instructions that define how your agent should behave, respond, and approach tasks

Why it matters: Rules are the most important part of creating a successful agent. They determine your agent's personality, tone, level of detail, decision-making approach, and how it uses connected tools and datasources. Well-written rules prevent unwanted behaviors and ensure consistent, high-quality outputs.

Best practice: Be specific and comprehensive. Include:

See our documentation on writing good agent rules for detailed guidance.

This section grounds your agent in your organization's specific knowledge, dramatically improving accuracy and relevance.

What it is: A toggle to enable user context for your agent

Why it matters: The key decision in this section is whether or not to click the Enable User Context button. When enabled, your agent can access information about the user it's conversing with, such as their name, role, or other profile information. This allows for more personalized interactions.

Impact on success:

Best practice: Enable user context when personalization improves the user experience, such as for HR agents, onboarding assistants, or role-specific tools. Disable it when user information isn't relevant to the agent's purpose.

What it is: Connected files, databases, or websites that your agent can reference when responding

Why it matters: Datasources transform your agent from a general-purpose AI into a specialized expert on your specific information. They enable your agent to answer questions with your company's actual data, policies, and documentation rather than relying on general knowledge.

Best practice: Connect datasources that are directly relevant to your agent's purpose. More datasources aren't always better—focus on quality and relevance. For more information, refer to our Datasources article.

Tools extend your agent's capabilities beyond conversation, enabling it to perform actions like searching the web, creating documents, analyzing data, or generating images.

Why it matters: The right tools enable your agent to complete tasks that require more than just text responses. However, unnecessary tools can confuse the agent or lead to unwanted behaviors.

Available tools include:

Allows your agent to search Google and retrieve current information from websites.

Use when: Your agent needs access to current information, public data, or content not available in your datasources.

Enables your agent to create and edit text documents and tables in an iterative workspace.

Use when: Users will work with your agent to create, refine, or iterate on written content like emails, reports, or documentation.

Provides the ability to analyze spreadsheets, run calculations, and query structured data.

Use when: Your agent needs to perform calculations, analyze datasets, or work with numerical data.

Allows your agent to create images based on text descriptions.

Use when: Your agent needs to generate visual content, diagrams, or illustrations.

Provides advanced capabilities for complex file generation, data processing, and specialized tasks beyond standard tools.

Use when: Your agent needs to create professional documents (PDFs, presentations), process large datasets, or perform advanced operations that exceed the capabilities of standard tools.

Enables your agent to perform advanced search across connected datasources with enhanced relevance and context understanding.

Use when: Your agent needs to find specific information across multiple datasources or when search accuracy and relevance are critical to the agent's purpose.

Enables your agent to provide references and citations for information pulled from datasources.

Use when: Accuracy and traceability are critical, such as for compliance, research, or policy-related agents.

This section shapes the user's first impression and helps them understand how to interact with your agent effectively.

What it is: The initial message users see when they start a conversation

Why it matters: The welcome message sets the tone and frames the conversation. It's your opportunity to orient new users and set clear expectations about what the agent can help with.

Best practice: Use the welcome message to:

Default message: "How can I help you right now?"

Better example: "I'm here to help you create sales proposals. Share your client's name and requirements, and I'll draft a customized proposal using our latest pricing and product information."

What it is: Sample prompts or tasks that demonstrate what users can ask

Why it matters: Example tasks reduce the learning curve and help users understand the agent's capabilities through concrete examples.

Best practice: Provide 1-4 specific, realistic examples that represent common use cases.

Example:

These settings control who can access your agent and how sensitive information is handled.

What it is: Controls whether your agent is Private (limited access) or Public (available company-wide)

Why it matters: Visibility determines who can discover and use your agent.

What it is: User roles that control who can view or edit your agent

Available roles:

Why it matters: Proper permissions ensure that only authorized users can change your agent's behavior while still allowing broad usage.

These settings allow you and your organization to interact with your agent directly from Slack or Microsoft Teams, bringing AI capabilities into your team's existing workflow.

Why it matters: Slack and Teams integration meets users where they already work, increasing adoption and making your agent more accessible.

This is different than enabling your agent to post to Slack which would be set up in Actions.

Warning: These settings should only be changed if you are familiar with the AI provider's documentation and have a good understanding of the AI model's capabilities and limitations.

What it is: A customizable template that controls how your agent's settings are formatted into instructions for the AI model

Why it matters: The instruction template determines how your rules, context, and other settings are presented to the underlying AI model. Modifying this template can significantly impact your agent's behavior.

Technical details: Instructions use Jinja's template syntax. Available variables:

elvex validates your template before saving to ensure it's properly formatted.

What it is: The AI provider that powers your agent

Why it matters: Different models have different strengths, capabilities, and costs. Choosing the right model can improve performance for specific use cases.

Impact on success:

Best practice: Start with the default model unless you have specific requirements that another model addresses better.

Agent owners and users with Editor roles can edit agents at any time.

elvex tracks changes to agents over time and notes which version is currently in use. Each edit creates a new version, visible in the chat interface and logs.

When you edit an agent, all users are immediately moved to the newest version.

This includes:

Best practice: Test significant changes in a private copy before updating a widely-used agent. Consider communicating major changes to your users.

There are two ways to start a conversation with an agent: