MCP Websearch evolves

Clovis Web Search, launched earlier this year as an MCP server, takes another step forward. Previously focused on finding information, it now provides a complete web information workflow: discover, read, map, explore, and monitor.

The clovis-websearch server now exposes 5 complementary tools, designed to work together in a single workflow. No changes to your existing integrations are required: these tools are added to the MCP and remain available both in the chatbot and through the API.

The 5 tools

1. search — Web search

Runs web queries through the Tavily API, with privacy controls and advanced filters: date, domain, depth, and inclusion of images or raw content.

🔹 Main use : quickly find specific information, monitor news, or target specific sources without irrelevant web results.

2. extract — Content extraction

Retrieves the full text of one or more supplied URLs in markdown or text format. An advanced mode is available for protected or complex pages.

🔹 Main use : read or analyze the content of a specific link (article, report, technical page) while removing unnecessary HTML/CSS/JS.

3. crawl — Website exploration

Starts from a root URL and follows internal links up to a defined depth and number of pages, returning the text content of the pages found.

🔹 Main use : scan an entire website, extract a set of related pages (documentation, blog), or populate a local knowledge base.

4. map — Structure mapping

Similar to crawling, but only returns a list of discovered URLs without extracting text.

🔹 Main use : audit a website structure, generate a navigation map, list all pages on a domain, or prepare targeted extraction.

5. search_status — Service status

Checks the availability and health of the search service, the Tavily API, and security filters (guardrails).

🔹 Main use : monitoring, debugging, or technical checks before running critical tasks that depend on web search.

A logical web information workflow

These 5 tools are designed to work together as a consistent workflow, from discovery to large-scale collection.

search → discover URLs or information

extract → read the content of a known URL in detail

map → inventory a website structure before taking action

crawl → collect the content of a section or an entire website

search_status → check service reliability before proceeding

In practice, an agent can now choose the appropriate sequence on its own: check service status, find sources, map a website, and then extract relevant content, all without complex client-side orchestration.

Availability

These tools are available immediately on the clovis-websearch MCP server, both in the chatbot and through the API. Existing integrations do not need to change: the new tools are added alongside those already exposed.

To learn more, see the WebSearch documentation.

This is just one step

Web Search will continue to evolve with more advanced capabilities (DeepSearch, analysis across multiple sources, intelligent navigation…). These 5 tools lay the foundations for a web workflow that is more precise, more controlled, and easier to automate.