Skip to main content

Web Search & Page Fetch

Nibchat includes two built-in tools that let the agent browse the web — no MCP server required.

Enabling​

Set the API keys in server/.env:

# Web search via Tavily
TAVILY_API_KEY=tvly-...

# Page fetching via Jina Reader
JINA_API_KEY=jina_...

Each tool is enabled independently — you can use one without the other.

Powered by Tavily, this tool lets the agent search the web and return summarized results.

When TAVILY_API_KEY is set, the agent gains access to web_search and will use it automatically when it needs up-to-date or web-specific information.

Example prompt: "What's the latest news on the EU AI Act?"

fetch_page​

Powered by Jina Reader, this tool fetches a URL and returns the page content as clean markdown — stripping ads, navigation, and other noise.

When JINA_API_KEY is set, the agent gains access to fetch_page.

Example prompt: "Summarize this article for me: https://example.com/article"

Content size limit​

Large pages are truncated to avoid exceeding the model's context window. Control the limit with:

WEB_FETCH_MAX_TOKENS=2000

The value is in tokens (converted to approximate characters internally). Default is 2000.

Tool call visibility​

Unlike memory tools, web_search and fetch_page are visible in the chat UI. Users see the tool being called and the result returned, similar to any MCP tool call.