cloro tool for Open WebUI

Give Open WebUI models four tools from cloro:

Tool	What it returns	cloro credits
google_search	Google organic results, the AI Overview and the sources it cites, People also ask	7
ask_chatgpt	The answer that the ChatGPT web app shows, with its sources	7
ask_google_ai_mode	The answer that Google AI Mode shows, with its sources	6
ask_gemini	The answer that the Gemini web app shows, with its sources	6

Each source is also sent to the chat as a citation. The free plan has 500 credits a month.

Use it to see what Google and the AI assistants tell users about a topic or a brand, and which pages they cite. For plain web search, a SERP-only provider costs less.

There are two ways to connect cloro to Open WebUI. Option 1 is this tool. Option 2 is cloro's hosted MCP server, which needs no code.

Option 1: install the tool

1. Get an API key at dashboard.cloro.dev/api-keys.
2. In Open WebUI, open Workspace > Tools and select Create. Paste the contents of cloro_tool.py and save. You can also get it from the Open WebUI community site, or use Import from URL with the raw URL of cloro_tool.py.
3. Select Valves (the gear icon on the tool's row). Set:
   • api_key: your cloro API key.
   • country: the default country, as an ISO 3166-1 alpha-2 code (default US). The model can set a different country for each call.
4. Turn the tool on for a model (Workspace > Models, edit the model, Tools) or for one chat (the Integrations menu in the message input).

Example prompts:

• "What does ChatGPT recommend for best CRM for startups, and which sources does it cite?"
• "Does Google's AI Overview for 'ai visibility tracking' cite example.com? Check in the US and Germany."
• "Compare the Gemini and Google AI Mode answers for 'best running shoes for flat feet'."

The tool needs httpx and pydantic, which Open WebUI already installs.

Option 2: connect the hosted MCP server

Open WebUI supports MCP servers over Streamable HTTP from version 0.6.31. cloro's server is at https://mcp.cloro.dev/mcp.

1. As an admin, open Admin Panel > Settings > Integrations.

2. Under Tools > External Tool Servers, select + (Add Connection).

3. Select the Type value until it shows MCP Streamable HTTP.

4. Set URL to https://mcp.cloro.dev/mcp.

5. Set Auth to Bearer and paste your cloro API key in the key field.

6. Set Connection ID and Name, for example cloro.

7. Open Advanced and set Function Name Filter List to:

   scrape_google, scrape_chatgpt, scrape_google_ai_mode, scrape_gemini, resolve_google_redirect, list_countries
   

   The server also lists tools for other engines. Use only the tools in this list. The filter hides the others from the model.

8. Set Access Control for the users who can use the server, then save.

9. Turn the server on for a model or a chat, as in step 4 of Option 1.

Each MCP tool call is a sync request. It costs the engine's credits plus a 2-credit sync surcharge (see providers and pricing). list_countries is free and resolve_google_redirect costs 1 credit.

Open WebUI shows a warning that its MCP support is experimental. If the model does not call the tools, use a model with reliable native tool calling.

Tests

The tests stub HTTP. They do not call the cloro API.

uv run --with pytest --with httpx --with pydantic pytest -q

Set CLORO_CAPTURE_DIR to a folder of saved cloro responses (google.json, chatgpt.json, aimode.json, gemini.json, badkey.json, ...) to also parse those files.

License

MIT