web_search
Search the web for real-time information about any topic. Use this tool when you need up-to-date information that might not be available in your training data, or when you need to verify current facts.
LABSCO SUMMARY
Anthropic's search, available to clients that aren't Anthropic's.
The useful part is the domain filtering: an allow list turns open web search into a search over sources you have already decided to trust, which is what makes results usable in a research or compliance context. It bills against your own Anthropic key, so search volume shows up on that account rather than a separate search vendor's.
What it is
A single-purpose MCP server that exposes Claude's web search API as a tool, so any MCP client — not just Claude's own apps — can reach current information from the web through one interface.
What you get
- A web search run against Anthropic's web search API and returned to the calling model
- Domain control on each search:
allowedDomainsrestricts results to sites you name,blockedDomainsexcludes them - A cap on how many results come back per search, set with
maxResults, defaulting to 5
Requirements
An Anthropic API key with web search enabled, supplied as ANTHROPIC_API_KEY. Node.js 18 or higher. The package is claude-search-mcp (1.0.0); the documented path is a clone plus npm install, npm run build and npm link, which puts the mcp-server-claude-search command on your PATH for the client to launch. Set the key in the server's env block rather than relying on the fallback: without it the server looks for a config file at a fixed path under your home directory, which will not exist on a fresh machine.
Setup effort
Build from source — git clone https://github.com/Doriandarko/claude-search-mcp.git
USE CASES
01
Answering from current information instead of training data
Search the web mid-conversation and let the model read the results, for questions where anything older than today is wrong.
02
Restricting research to sources you trust
Name the domains that may appear in results — documentation sites, a regulator, your own site — so answers cannot be assembled from anything else.
03
Keeping known-bad sites out of an answer
Block domains you do not want quoted, such as content farms or a competitor, and run the search normally otherwise.
04
Adding web search to a non-Claude client
Point any MCP-capable client at this server to get search through your Anthropic key, without that client needing its own search integration.
1 tool
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
BEFORE YOU INSTALL
What to know first
Tools you get
1 tool when we connected.
Requires
- Node.js 18 or later
- ANTHROPIC_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
CONTEXT COST
Light Standing cost
Rated by the size of the one tool definition that rides along on every turn
INSTALLATION
Connect your client
This one is built from source rather than installed with a single command.
Server setup
TERMINAL
git clone https://github.com/Doriandarko/claude-search-mcp.git
cd claude-search-mcp
npm install
npm run build
npm linkMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"claude-search": {
"command": "mcp-server-claude-search",
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}npm link puts mcp-server-claude-search on your PATH. If Claude Desktop cannot find it, use the full path that which mcp-server-claude-search prints. <YOUR_ANTHROPIC_API_KEY> is an Anthropic API key that has web search enabled.
