Creating an agent in LibreChat

Questions

  • How do I create a custom AI agent in LibreChat?
  • What is the Agent Builder in LibreChat?
  • How do I create an AI assistant with a specific purpose?
  • How do I give an agent access to my own documents?
  • How do I access an agent I have already created?
  • Why is the Create button greyed out in the Agent Builder?

Environment

This article applies to all current Bowdoin faculty, staff, and students with access to LibreChat. Agent creation is available through the Agent Builder in the right-hand panel of the LibreChat interface. You will need a clear idea of what you want your agent to do and, optionally, any documents you want it to reference.

  • Who: All current Bowdoin LibreChat users
  • What: LibreChat Agent Builder at librechat.bowdoin.edu
  • Not covered: Using pre-configured shared Agents — see Choosing an AI model in LibreChat in the Related Articles section

Resolution

An AI agent in LibreChat is a customised version of an AI model that you configure with a specific name, instructions, capabilities, and tools. Agents can automate repetitive tasks, provide specialised assistance, and draw on your own uploaded documents to give more accurate, contextual responses.

Step 1 — Open the Agent Builder

  1. In the LibreChat window, click the panel icon in the upper-right corner of the interface to open the right-hand panel.
    LibreChat interface with arrows pointing to the right menu display icon and the Agent Builder button.
  2. From the panel drop-down, select Agent Builder.
Tip: If the panel is already open, look for the Agent Builder option in the drop-down at the top of the right panel.

Step 2 — Fill in your agent's basic information

The Agent Builder interface in LibreChat showing fields for name, description, category, instructions, and model.

  • Name (required) — Enter a clear, descriptive name that reflects the agent's purpose, such as "Research Assistant" or "Email Drafter".
  • Description (optional) — A short summary of what the agent does. Helpful if you plan to share the agent with others.
  • Category (required) — Organises agents within the platform. The default is "General", which works for most use cases.
  • Instructions (required) — The most important field. Write clear, specific instructions that define:
    • The agent's role and purpose — for example, "You are a helpful research assistant for Bowdoin faculty."
    • The tone and style to use — formal, conversational, concise, detailed
    • Specific tasks the agent should focus on
    • Any limitations or topics to avoid
  • Model (required) — Select the AI model that will power your agent. GPT-5 is the default and suitable for most tasks.
Tip: Think of the Instructions field as a system prompt. The more specific and direct you are, the more consistent your agent will be. Starting with "You are..." and using clear directives works well. You do not need to be polite — direct instructions produce better results.

Step 3 — Configure capabilities

The Capabilities section in the Agent Builder showing the File Context, Artifacts, and File Search options.

  • File Context — Upload reference documents (PDFs, guides, policies, course materials) that your agent will always have access to when responding. Note: you can only upload files after the agent has been created.
  • Artifacts — Enables the agent to produce structured outputs such as charts, code blocks, or formatted documents in a separate interactive panel. Toggle on if your agent should generate visual or structured content.
  • File Search — Allows the agent to search across a set of uploaded documents to answer questions. Useful for agents that serve as a knowledge base or policy library.

Step 4 — Add Tools, Actions, and Support Contact

Tools and Actions, Support Contact, and Create button with a red arrow pointing to the Create button in the Agent Builder.

  • Tools (advanced) — Connect pre-built tools such as web search or API connectors. Optional for most users — can be added later by editing the agent.
  • Actions (advanced) — Create custom actions that let the agent interact with external services via APIs. For users who need the agent to connect to external systems.
  • Support Contact (optional) — Add a name and email address if you plan to share the agent with a team or department, so others know who to contact with questions.

Step 5 — Create your agent

  1. Scroll through the form to review your settings — especially the Instructions field.
  2. Click the green Create button at the bottom of the form. Your agent is saved immediately.
  3. If you set up File Context or File Search, return to the agent's settings after creation to upload your documents.
  4. Select the agent from the model drop-down in the upper-left corner of the chat window to start a conversation and test it.
Tip: If your agent does not behave as expected, edit it and refine the Instructions field. Small changes to wording can have a significant impact on responses. Iteration is normal.

Step 6 — Access your agents

All agents you have created (and any shared with you) are available under My Agents in the model selection drop-down in the upper-left corner of the chat window.

The My Agents section in the LibreChat model selection menu.

Tips for writing effective agent instructions

  • Be specific about the agent's role: instead of "You are a helpful assistant", try "You are a writing coach who helps Bowdoin students improve their academic essays."
  • Set the tone explicitly — formal, conversational, concise, or detailed
  • Define scope — what the agent should focus on and what it should avoid
  • Include examples of the style of response you expect
  • Create, test, then refine — small iterations produce the best results

Troubleshooting

  • The Create button is greyed out — make sure all required fields (marked with an asterisk) are filled in: Name, Category, and Model are all required.
  • My agent isn't following the instructions — rewrite the Instructions field to be more direct and specific. Avoid vague language; starting with "You are..." and using clear directives works best.
  • I can't upload files for File Context or File Search — files can only be uploaded after the agent has been created. Create the agent first, then return to its settings to add documents.
  • I need help with advanced features (Tools, Actions, Custom Prompt Mode) — submit a ticket through the Service Catalog and a member of the IT team will follow up.

Additional Help

If you need further assistance, you have several options:

  • Bowdoin Bot: Chat with Bowdoin Bot directly from any KB page for instant answers.
  • Phone: Call the Bowdoin College Service Desk at (207) 725-3030.
  • In person: Visit the Tech Hub in Smith Union during business hours.
  • Submit a ticket: Request assistance through the Service Catalog.
Print Article

Related Articles (4)

Guidance on choosing the right AI model or Agent in LibreChat, including a comparison table of available model groups (OpenAI GPT, Anthropic Claude, Amazon Nova, Meta LLaMA, DeepSeek), when to use each, and how to access pre-configured Agents such as the Image Generation Agent and Artifacts Agent.
Instructions for logging in to LibreChat, Bowdoin's AI chat platform, for the first time — including both sign-in methods, completing the Terms of Service, and an overview of the interface layout. Applies to all current Bowdoin faculty, staff, and students with LibreChat access.
Tips and techniques for writing clear, effective prompts to get better results from AI tools including LibreChat, Copilot, and Bowdoin Bot. Covers the basics of prompt engineering with practical examples. Applies to all Bowdoin students, faculty, and staff.
Instructions for uploading files to LibreChat so the AI can reference their contents, managing uploaded files, enabling File Search for agents, and using the Artifacts feature to generate and preview structured outputs such as code, diagrams, and HTML in a separate interactive panel.