JcDizzy

JcDizzy / CC-Web-MCP

Public

面向 Claude Code 第三方模型后端的轻量本地 WebSearch/WebFetch fallback MCP。A lightweight, local-first WebSearch/WebFetch fallback MCP for Claude Code when using third-party Anthropic-compatible models such as DeepSeek, Qwen or Kimi.

14
1
89% credibility
Found May 20, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

CC Web MCP is a free tool that gives AI coding assistants the ability to search the internet and read web pages. It was created specifically for people who use third-party AI models (like DeepSeek, Qwen, or Kimi) with Claude Code, because those models don't have built-in web browsing features like official Claude models do. The tool automatically intercepts web requests and routes them through itself, while keeping official Claude models on their native tools. It includes safety features to prevent accessing private networks, and can be set up with a single command.

How It Works

1
💭 You discover Claude Code needs web access

You realize your favorite AI coding assistant can't browse the web when you use DeepSeek, Qwen, or Kimi instead of Claude.

2
🔌 You find a tool that fills the gap

You discover CC Web MCP, a lightweight tool that gives your AI assistant the ability to search and read web pages.

3
🚀 You set up the connection in one click

You run a simple command to connect the tool to your AI assistant, and it automatically configures everything for you.

4
✨ Your AI assistant learns to browse

Now when you ask your AI about current topics, it can search the web, read articles, and gather information just like a human would.

5
Different models, smart routing
🤖
Third-party models (DeepSeek, Qwen, Kimi)

Automatically routes through CC Web MCP for web access

🧠
Official Claude models

Continues using their native web search and reading tools

6
🔒 Safety features keep you protected

The tool blocks access to private networks and sensitive addresses, so your AI can only browse public web pages safely.

🎉 Your AI assistant can now browse the web!

You can ask your AI about anything, and it will search, read, and summarize web content to help you with your projects.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is CC-Web-MCP?

CC-Web-MCP is a Python-based MCP server that gives Claude Code web search and web fetch capabilities when using third-party Anthropic-compatible models like DeepSeek, Qwen, or Kimi. These models lack native web access, so this tool fills the gap by providing a local fallback layer. It exposes three main tools: web_search for querying public pages, fetch_url for converting URLs to markdown, and research_brief for lightweight multi-source summaries. The server runs over stdio and integrates directly with Claude Code's hook system to route web requests intelligently based on which model is active.

Why is it gaining traction?

The hook-based routing is the standout feature. Instead of a blunt "use this tool or don't," it intelligently detects whether Claude Code is running an official model or a third-party one, then routes WebFetch calls accordingly. This means official Claude models keep using native capabilities while DeepSeek and friends get cc-web instead. The search backend fallback chain (DuckDuckGo to Bing to Bing CN) with automatic cooldown and retry logic handles reliability without user intervention. Security boundaries that block localhost, private IPs, and cloud metadata endpoints prevent accidental data leakage.

Who should use this?

Developers running Claude Code with DeepSeek, Qwen, Kimi, or similar third-party models who need web research capabilities. If you've been manually copying search results or giving up on tasks that require current information, this eliminates that friction. Teams evaluating cost-effective alternatives to Claude's official models will find the most value here.

Verdict

A genuinely useful niche tool for a specific workflow. The credibility score of 0.8999999761581421% reflects a very small community (14 stars), and the project is marked alpha. That said, the documentation is thorough and the doctor command makes troubleshooting straightforward. If you fit the use case, this is worth the setup time. If you're using official Claude models, skip it.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.