Give Open WebUI models live web data from MrScraper: retrieve public pages as Markdown or HTML, search Google, and check your MrScraper plan and token usage.
There are two ways to connect. Use the MCP server if you are an Open WebUI admin on v0.6.31 or later; use the community tool otherwise.
- Open Admin Settings → External Tools and click +.
- Set Type to MCP (Streamable HTTP).
- Set Server URL to
https://mcp.mrscraper.com/mcp. - Choose an Auth method:
- Bearer, with a MrScraper API key from app.mrscraper.com/api-tokens as the key; or
- OAuth 2.1, which registers Open WebUI with MrScraper through dynamic client registration and asks each user to sign in to MrScraper.
- Save, then enable the server for the models that should use it.
The server provides scraperAPI, webSearch, and status. See the
MCP server documentation.
- Open Workspace → Tools and click + (or import the tool from the Open WebUI community site).
- Paste the contents of
mrscraper_tool.pyand save. Open WebUI installs its dependency,mrscraper-sdk. - Open the tool's Valves and set
api_keyto your MrScraper API key. - Enable the tool in a chat or on a model.
| Function | What it does |
|---|---|
scrape_page |
Retrieves a public page as Markdown (default) or HTML. Long pages are cut at max_characters (Valve, default 100,000; 0 disables). |
web_search |
Searches Google and returns organic results (rank, title, link, description, source) and the next page number. |
account_status |
Shows the subscription status, token limit, tokens used, and tokens remaining. Uses no tokens. |
Errors are reported to the model as short messages with the HTTP status; the API key never appears in them.
Target URLs, search queries, and request options are sent to MrScraper's API,
and scrape_page and web_search use tokens from your MrScraper plan. Use
them only for public or otherwise authorized content. See the
Privacy Policy and
Acceptable Use Policy.
Support: support@mrscraper.com