Skip to content

LibreChat

Jump to bottom
Parker Combs edited this page Oct 17, 2025 · 7 revisions

About

LibreChat is an open-source, ChatGPT-like graphical interface for interacting with LLMs. Available features include:

  • Having persistent conversations with an AI model
  • Uploading documents for a model to interpret
  • Creating persistent AI agents specialized for a specific task
  • Generating tools for an agent from an API

The NIEHS deployment of LibreChat is available here.

The full LibreChat documentation is available here.

The LibreChat GitHub repository is available here.

Agents

Adding tools to an agent

When creating an agent, there are multiple ways to add tools for the agent to call: adding tools to an agent

  1. Add MCP Server Tools - Add tools from available MCP servers.
  2. Add Tools - Add tools from sources provided by LibreChat.
  3. Add Actions - Provide an OpenAPI schema to automatically generate tools.

For methods 1 and 2, simply click on the corresponding button to open a modal containing available tools to add to the agent. Click the green "Add+" button for each tool you would like to add. For method 3, follow the below instructions:

  1. Click the Add Actions button. You will be taken to a new screen to fill in details about the API you would like to generate tools from.
  2. Click the "Authentication" dropdown menu (if applicable). Select the correct authentication method and fill in your credentials (e.g., an API key).
  3. Retrieve then OpenAPI schema JSON or YAML from the API you would like to use.
  4. Under the "Schema" header, copy and paste your retrieved OpenAPI schema in the corresponding text box. If the schema is valid, a set of "Available Actions" should appear under the schema text box, with one entry per endpoint in the API.

agent actions

  1. Supply the URL to the API's privacy policy in the text box under the "Privacy Policy URL" header, if applicable.
  2. Press the green "Create" button.

Uploading files with agents

You may upload a file to an agent to be used in a file search.

  1. Select an agent or create a new agent.
  2. Ensure the Enable File Search option is checked, and click Upload for File Search and select one or more files to upload.

agent upload file

  1. The agent should now have knowledge of this document. You may query the LLM, and it will conduct a file search to find relevant information across its known files to respond to the query. Keep in mind that agents do not retain knowledge of the structure of the document, only its content; therefore, queries pertaining to specific pages or sections will not be understood correctly by the model.

querying agent

Clone this wiki locally