Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
To use a web-capable MCP server with Claude Code, add the server using the endpoint or launch command in that provider’s setup guide, check its connection in /mcp, then ask Claude to use one of its tools. “Web MCP” is not one universal service: search, page retrieval, and browser control may come from different servers, with different transports, credentials, and tools.
Anthropic’s official examples illustrate two distinct things: a hosted server for searching Claude Code documentation and a local Playwright server for browser automation. Neither is a general-purpose web search provider. This guide shows how to configure those examples, adapt the steps to another server without guessing its details, and tell whether Claude actually used it.
What “Web MCP” means in Claude Code
MCP servers give Claude Code tools beyond its built-in set. A web-capable server might search the web, retrieve pages, or control a browser; what it can do depends on the server’s actual tools. Claude Code can connect to hosted servers and launch local processes. Anthropic describes MCP as a way for Claude Code to use tools such as an issue tracker, database, or web browser in its official quickstart.
Free tools Windows power users keep installed
One-click scans. No signup required.
Before installing anything, decide which capability you need. Search finds candidate pages; retrieval reads a page; browser automation interacts with a site. A server may provide one or combine several, but do not assume it searches the public web, can read every URL, returns citations, or has fresh results unless its provider documents those behaviors.
#1 Best Overall
The endpoint, transport, authentication, environment variables, tool names, and terms are provider-specific. The official Claude Code docs endpoint below searches Claude Code documentation only. For general web search, follow the selected provider’s current instructions instead of reusing that URL.
Choose hosted HTTP or local stdio
| Choice | Where it runs | Configuration shape | Official example | What to check |
|---|---|---|---|---|
| Hosted HTTP | A remote service | --transport http plus the service URL |
Claude Code documentation search | Network access and any required service authentication |
| Local stdio | A process launched on your computer by Claude Code | A command after -- |
Playwright browser automation | Local runtime, package, and browser prerequisites |
Claude Code’s MCP reference documents supported connection methods. Choose the transport named in the provider’s current installation guide; do not change a provider’s HTTP, SSE, stdio, or other supported remote setup into a different transport by guesswork.
How do I add a web MCP server to Claude Code?
1. Read the server provider’s setup instructions
Find the provider’s official install instructions and record its exact URL or process command, required transport, authentication steps, and environment variables. Determine whether the server is a search tool, page reader, browser controller, or some combination. If it uses OAuth, use the sign-in flow it documents. If it requires a token, configure it only using the documented method; do not paste a made-up token flag into a command.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →2. Add a hosted HTTP server
Anthropic’s hosted example adds the Claude Code documentation search server:
claude mcp add --transport http claude-code-docs https://code.claude.com/docs/mcpFor another hosted server, replace claude-code-docs with a name you choose and replace the URL with the provider’s exact endpoint. Keep --transport http only if that provider specifies HTTP. If the server requires OAuth, complete authorization through Claude Code’s /mcp interface when prompted.
This example is useful for looking up Claude Code documentation; it is not a general search endpoint. Do not treat its successful connection as proof that Claude can search arbitrary websites.
3. Add a local stdio server
The official quickstart’s Playwright example launches a local browser-automation server:
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →claude mcp add playwright -- npx -y @playwright/mcp@latestThe -- separates Claude Code’s options from the local process command. The quickstart states that this Playwright example requires Node.js 18 or later. It demonstrates browser automation, not a general web-search service. Follow the current Playwright and provider instructions for any additional browser prerequisites.
Choose where the server configuration applies
Claude Code supports three scopes. The default local scope is private to the current project; user makes the server available to you across projects; project stores configuration in the project root’s .mcp.json so it can be shared with teammates, subject to their approval.
| Scope | Where it is stored | When to use it |
|---|---|---|
local (default) |
In ~/.claude.json under the current project’s entry |
You need the server only for this project and do not need to share it. |
user |
Top-level mcpServers entry in ~/.claude.json |
You want the server available across your projects. |
project |
.mcp.json in the project root |
You want to share project configuration with teammates. |
To add the hosted documentation example at user scope, use:
Rank #3
claude mcp add --scope user --transport http claude-code-docs https://code.claude.com/docs/mcpProject-scoped servers require user approval. Review the server command or URL and any exposed settings before adding .mcp.json to version control. Claude Code reads that file at session start, so restart the session after editing it.
How can I check whether an MCP server is connected?
- Check the CLI configuration: in a terminal, run
claude mcp list. Confirm the server appears under the name you added. - Inspect the session connection: start Claude Code and enter
/mcp. Check the server’s connection status and, where available, its discovered tools and authentication state. - Test a real tool call: ask Claude to do a task that clearly requires that server. For the documentation example, you could say:
Use the claude-code-docs server to look up how MCP server scopes work. - Confirm which tool ran: check Claude’s output for a tool call labeled with your server’s name. If another built-in capability could answer the same question, naming the server in the prompt makes this check more useful.
A listed server alone does not establish that its tools are usable. Check the connection state and tool list, then verify an actual call. A server can be configured but unauthenticated, fail to connect, lack registered tools, or be waiting for project approval.
How do I make Claude use the web MCP server?
State the task and, when useful, the server name. For a provider that documents a web-search tool, ask Claude to search for a specific subject and use that server. For a page-reading tool, give it a URL and ask it to retrieve or summarize that page. For browser automation, describe the interaction you want Claude to perform. These prompts are examples of how to direct a request, not guarantees about a server’s capabilities.
If Claude answers without calling the expected server, make the required capability explicit and check the tool call display. Then confirm in /mcp that the server is connected and exposes the relevant tool. Tool names and behaviors come from the selected server, so use its documentation rather than assuming every provider calls a tool “search” or “fetch.”
Do not promise yourself citations, coverage of a particular search engine, access to pages behind a login, or a particular freshness level unless the provider states that the service supports it. A browser-control server may be able to interact with a page without offering web search; a search server may return results without retrieving each full page.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhat MCP does not configure: Anthropic’s platform web tools
Anthropic’s Platform documentation separately describes web search and web fetch tools, including versioned tool types, in its tool reference. That is distinct from adding an external server to the Claude Code CLI. The platform’s built-in/API tool reference does not establish that a general-purpose web-search MCP server is bundled with Claude Code.
Likewise, the MCP connector for the Anthropic API is a separate configuration path. Its API settings, including mcp_servers and mcp_toolset, are not Claude Code CLI commands; the API connector documentation says its server must be publicly exposed via HTTP. Use the instructions for the product and environment you are configuring.
Or skip the browser setup
If your immediate goal is to capture a webpage rather than search or ask Claude to read it, ScreenshotNeo is a screenshot API and MCP server for developers. It is not a general web-search endpoint or a replacement for a page-reading server. Its capture options include PNG, JPEG, WebP, or PDF output, and its MCP server provides take_screenshot, get_page_info, and capture_pdf tools for Claude, Cursor, and other MCP clients.
For the browser-independent API call, install Python’s requests package first. Use your own API key in place of the example value. The ScreenshotNeo API documentation describes its API and options.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
cURL
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpPython
import requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)Node.js
const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);ScreenshotNeo removes known consent banners, newsletter popups, and chat widgets before capture; each of those steps can be turned off. Bot checks, blank pages, timeouts, failed loads, and cache hits cost nothing, and responses report the page verdict and billing status in headers. It also offers an MCP server for AI agents. The free plan includes 1,000 screenshots per month with no card; paid plans start at $5 for 3,000 shots. Sign up for ScreenshotNeo’s free plan.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Troubleshooting Claude Code MCP connections
No MCP servers configured
Confirm you ran claude mcp add in the intended project and selected the scope you wanted. Claude Code’s documented configuration locations include ~/.claude.json and the project-root .mcp.json; a server saved under one project or scope will not necessarily appear in another. Check claude mcp list from the expected working context.
Connection fails
Run claude mcp list and claude mcp get <name> to inspect the configured server. Verify the provider’s exact URL or process command, the specified transport, network reachability, required credentials, and any documented environment variables. For a project-scoped server, check whether approval is pending in /mcp.
It connects, but no tools appear
Open /mcp and inspect the tools registered for that server. The official troubleshooting guidance notes that missing environment variables, such as an API key, can prevent a server from registering tools. Confirm that required variables are available to the process in the way the provider specifies, then reconnect or restart Claude Code if needed.
Edits to .mcp.json have no effect
Start a new Claude Code session: the project configuration file is read at session start. Also check that the file is in the project root and contains the server entry you intend to share.
A local server starts too slowly
The Claude Code quickstart documents a 30-second default startup timeout and the MCP_TIMEOUT environment variable, measured in milliseconds, as a way to increase it. If startup exceeds the default, check the current official quickstart troubleshooting guidance for the supported setting and apply it to the environment that launches Claude Code. Runtime behavior can change, so use the live documentation if the setting does not work as expected.
Claude replies but does not call the server
Ask for a task that requires the server’s documented capability, name the server, and inspect the tool-call label. If no tool is available, verify the connection and tool list in /mcp. A search request will not necessarily trigger a browser-control tool, and a browser-control server will not necessarily provide search.
Sources and scope
The Claude Code setup and troubleshooting details above follow Anthropic’s official MCP quickstart and MCP reference. The platform/API distinction follows Anthropic’s tool reference and MCP connector. These pages were retrieved on September 29, 2026; the pages reviewed do not state publication dates. Provider endpoints, authentication flows, availability, and tool behavior can change, so use the selected provider’s current setup instructions for its own service.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
