CORAS

Create and Configure an AI Agent

Build a custom AI agent tailored to a specific mission role, communication style, and data context.

AI Agents are custom AI personas you configure to perform specific tasks, follow defined instructions, and communicate in a chosen style. Each agent can be scoped to specific lists, connected to validated prompts, tested before deployment, and exported for reuse across tenants. By customizing how an agent behaves and communicates, you align AI outputs with your team's standards and mission requirements.


Access AI Agent Settings

  1. Go to the top right corner of the screen

  2. Click the dropdown arrow next to your name

  3. Select AI Settings

  4. In the left-hand menu, click AI Agent


Create a New AI Agent

  1. Click + Add Agent (top right of the page)

  2. Complete all required fields:


Required Fields (Must Be Completed)

  • Agent Name - Enter a clear name for the agent
    Example: Chart Specialist

  • Agent Description - Briefly explain what the agent does
    Example: Helps analyze charts and explain data trends

  • Communication Style - Select how the agent should communicate:

    • Formal – Structured, serious, and professional tone; avoids casual language

    • Casual – Relaxed and conversational; more informal and easygoing

    • Professional – Clear, concise, and business-appropriate tone

    • Friendly – Warm, approachable, and supportive tone

    • Technical – Detailed and precise; uses domain-specific or technical language

  • Role - Define the agent’s role
    Example: Analyze data and provide insights

  • Instructions - Enter a detailed prompt that tells the AI how to behave

    Your instructions should include:

    • What the agent is responsible for

    • How it should respond (tone, format, level of detail)

    • Any rules or limitations

    Example:
    You are a Chart Specialist. Analyze charts and provide clear insights. Highlight trends, patterns, and anomalies. Use a professional tone and keep responses concise.

The Instructions field is the most important part of agent configuration. The more specific and mission-aligned your instructions, the more consistent and useful the agent's outputs will be.


Optional Fields - Under "Data Configuration" Tab

  • Agent Image - Upload and image to represent the AI Agent

  • Lists (Dropdown) - Select which list(s) this agent is available to run in. This setting is authoritative — it determines where the agent can run in chat regardless of what prompts are bundled into it.

    When a prompt is bundled into an agent, the prompt's own available-for-lists setting is ignored. Only the agent's list scope applies, so the agent's availability stays predictable no matter which prompts it bundles.

    One exception: if a bundled prompt has a required list that isn't in the current chat context, that specific prompt is blocked. The agent and its other prompts keep working. The same prompt used on its own, outside an agent, still honors its own available-for-lists setting.

  • Prompts (Dropdown) - Select prompt(s) from the Smart Prompts to associate with the agent. Linked prompts appear as options when users interact with the agent.

  • Kick-off Prompt (Dropdown) — Select a prompt to open automatically when any user starts a chat with this agent. The selected prompt pre-populates the chat composer before the user types anything — they can edit it or send it as-is. Leave blank for no kick-off prompt.

  • Files (Dropdown or Drag) - Attach source files for Gary to reference whenever a user chats with this agent. Files uploaded here are available in every conversation with this agent automatically — users don't need to upload them manually. Supported file types: PDF, Word (.docx), Excel, plain text (.txt), Markdown (.md), and JSON.


  1. Click Apply

  2. Click Save

Your AI Agent is now created and available for use.


Edit and Test an AI Agent

  1. Locate the AI Agent in the list

  2. Click the pencil icon next to the agent


When editing an agent, the Agent Testing panel will appear on the right side of the screen.

Test the agent:

  • Use the Ask me something text box to enter a question or request

  • Review the response and verify:

    • The agent is following its instructions

    • The tone matches the selected communication style

    • The outputs are accurate and useful for the intended use case

Always test before finalizing. A few test prompts will surface gaps in the instructions that aren't obvious until the agent is running.


Add Data Context

To give the agent access to specific data during testing:

  1. In the top right of the Agent Testing panel, click Configure Context

  2. Select a Node

  3. Select a List

  4. Click Save

Gary will use that data when generating responses in the testing panel.


Manage Agents

To access management options for any agent:

  1. Locate the agent

  1. Click the three vertical dots (⋮) between the Agent Name and Description

Available Options

  • Duplicate Agent - Create a copy of an existing agent configuration. Use this when you want to build a new agent based on an existing one without modifying the original.

  • Export Agent – Download the agent configuration as a JSON file for reuse in other tenants

  • Delete Agent – Remove the agent permanently


Export an AI Agent

  1. Click the three dots (⋮) next to the agent

  2. Select Export Agent

  • A JSON file will download to your computer

  • This file contains the full agent setup and can be reused


Import an AI Agent

  1. Click Import Agent (top right of the page, next to + Add Agent)

  2. Drag and drop the JSON file into the blue upload box, or select the file manually

  1. Click Continue

The agent is imported into your current tenant and is immediately available for use.


Example AI Agents

  • Chart Specialist - Explains charts and provides data insights

  • Contracting Requirements Analyst - Helps draft and review acquisition requirements — SOW sections, PWS criteria, and evaluation factors — using a formal, compliance-oriented tone aligned to FAR/DFARS standards.


Tips

  • Be clear and specific in the Instructions field

  • Choose the correct communication style for your use case

  • Use the Agent Testing panel before making an agent available to your team

  • Add data context if the agent needs access to specific information

  • Scope agents to lists when possible — it keeps outputs focused and relevant

  • Export agents before making major changes — you can always reimport a working version


Get Help

Contact your Customer Success Manager or visit the Customer Resource Center for additional support with agent configuration and AI Settings.