Nymbo/Tools MCP

πŸ”

Enable Image Gen, Video Gen, and Deep Research

The Generate_Image, Generate_Video, and Deep_Research tools require a HF_READ_TOKEN set as a secret or environment variable.

  • Duplicate this Space and add a HF token with model read access.
  • Or run locally with HF_READ_TOKEN in your environment.
These tools are hidden as MCP tools without authentication to keep tool lists tidy, but remain visible in the UI.
🧠

Persistent Memories and Files

In this public demo, memories and files created with the Memory_Manager and File_System are stored in the Space's running container and are cleared when the Space restarts. Content is visible to everyoneβ€”avoid personal data.

When running locally, memories are saved to memories.json at the repo root for privacy, and files are saved to the Tools/Filesystem directory on disk.

πŸ”—

Connecting from an MCP Client

This Space also runs as a Model Context Protocol (MCP) server. Point your client to:
https://mcp.nymbo.net/gradio_api/mcp/

Example client configuration:

{
  "mcpServers": {
    "nymbo-tools": {
      "url": "https://mcp.nymbo.net/gradio_api/mcp/"
    }
  }
}

Run the following commands in sequence to run the server locally:

git clone https://huggingface.co/spaces/Nymbo/Tools
cd Tools
python -m venv env
source env/bin/activate
pip install -r requirements.txt
python app.py
πŸ› οΈ

Tool Notes & Kokoro Voice Legend

No authentication required for: Web_Fetch, Web_Search, Code_Interpreter, Memory_Manager, Generate_Speech, File_System, and Shell_Command.

Kokoro voice prefixes

  • af β€” American female
  • am β€” American male
  • bf β€” British female
  • bm β€” British male
  • ef β€” European female
  • em β€” European male
  • hf β€” Hindi female
  • hm β€” Hindi male
  • if β€” Italian female
  • im β€” Italian male
  • jf β€” Japanese female
  • jm β€” Japanese male
  • pf β€” Portuguese female
  • pm β€” Portuguese male
  • zf β€” Chinese female
  • zm β€” Chinese male
  • ff β€” French female

Web Fetch

Convert any webpage to Markdown, inspect the raw HTML response, or extract all links. Supports custom element removal, length limits, and pagination with offset.
0 20000
0 100000
Mode

Markdown cleans content, HTML returns raw response, URL Scraper lists links.