• JSON
  • MARKDOWN
  • TEXT
  • HUMAN

How to get structured data from your AI agents

Connect LibreChat once, then tell your agent exactly how to format its answers. You'll get clean, predictable data instead of conversational filler.

Ask AI about this page

Short answer

How do I make my AI agent output specific data formats?

Use the open_responses tool to define a schema or a specific layout for your agent to follow. This forces the agent to deliver structured JSON, Markdown, or plain text without the usual conversational chatter. You'll see the specific outcomes below.

The results.

Where your data lands.

The agent processes your request and routes the output based on the format you requested.

JSON

Clean data objects

The agent wraps its response in a valid JSON object. This makes the data ready to be used by your other scripts or databases.

MARKDOWN

Formatted documents

The agent produces structured Markdown tables or headers. This is useful for reports you need to read or copy into documents.

TEXT

Strictly plain text

The agent strips away the conversational fluff. You get only the raw information you asked for without any introductory sentences.

HUMAN

Manual review needed

If the agent can't fit the data into your requested schema, it will flag the error. You'll need to step in and fix the prompt or the schema.

The workflow

What your AI does when a structured request arrives.

The agent shifts from conversational mode to data-generation mode.

  1. Format definition

    The agent receives your instruction and identifies the required structure, such as a specific JSON schema.

    open_responses

  2. Constraint application

    The agent applies the formatting rules to its internal reasoning process to ensure the output matches your template.

    open_responses

  3. Data extraction

    The agent pulls the necessary facts from the conversation or your files and maps them to the requested keys.

    open_responses

  4. Validation

    The agent checks the final string against the requested format to ensure it isn't broken before showing it to you.

    open_responses

Try it

Copy these to start.

Use these prompts to immediately change how your agent responds.

Starting points

Swap out the bracketed text for your actual data requirements or specific keys.

LibreChat Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_2Y5OcBZjfHKkJsGbCGI3uDOzGTusTgIZgBtbs3a3/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — LibreChat capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "librechat-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_2Y5OcBZjfHKkJsGbCGI3uDOzGTusTgIZgBtbs3a3/mcp"
    }
  }
}
Copy into chat04

  • Convert the summary of this GitHub repo into a JSON object with keys for 'stars', 'forks', and 'description'.

  • Take the last three messages in this Slack channel and format them as a Markdown table.

  • Extract all the email addresses from this contact list and return them as a raw JSON array.

  • Summarize this project README into a structured Markdown list with headers for 'Installation' and 'Usage'.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect LibreChat once, then ask.

Just link your account once. Your credentials stay encrypted, and you can start using these tools in your favorite client immediately.

Connect LibreChat to your AI

FAQ

How the structured data task behaves.

  • 01

    Can the AI change my existing files to JSON?

    No. The agent only generates the structured text in the chat window. It does not overwrite your local files.

  • 02

    Will the agent always follow the JSON schema?

    The agent tries to follow the schema strictly, but if the data is missing or the request is too vague, it might fail. You'll have to refine the prompt.

  • 03

    Can I use this to create new files automatically?

    No. This tool is for generating structured text responses within the chat. It does not have permission to write to your hard drive.

  • 04

    Does this work with any AI client?

    It works with any MCP-compatible client that you connect to the Vinkius platform.

  • 05

    Can the agent output images instead of text?

    No. The open_responses tool is specifically for text-based formats like JSON, Markdown, and plain text.

  • More questions about LibreChat? The Connector page answers them. See everything the LibreChat Connector can do

Connect LibreChat to Claude, Cursor, ChatGPT & more