Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@websearch-deepseek What are the latest AI industry news?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A universal MCP Server that provides a web search tool powered by DeepSeek's native web search API — no third-party search API required.
Works with Claude Code, Cursor, Continue, OpenCode, pi, and any tool that supports the MCP protocol.
🔍 DeepSeek Native Search — Uses the server-side web_search_20250305 tool, no third-party search API needed
📝 AI-Generated Answers — Returns detailed answers synthesized from full page content, not just a list of URLs
🔗 Source URLs Included — Every answer comes with original source links for verification
⚙️ Flexible Configuration — Choose model, toggle thinking mode, adjust token limits
🌐 MCP Protocol — Standard JSON-RPC over stdio, compatible with all MCP clients
Related MCP server: websearch-deepseek
User Query → DeepSeek Model
↓
Server executes web_search (fetches pages)
↓
Encrypted page content → Server decrypts → Feeds to model
↓
Model generates detailed answer based on full content
↓
Returns: AI-generated answer + source URL listOne MCP tool call = One DeepSeek API request — search, decryption, and answer generation all happen server-side.
Visit DeepSeek Platform to sign up and get your API Key.
npm install -g websearch-deepseekAdd the following to your MCP configuration file:
Edit ~/.claude/claude_desktop_config.json or .mcp.json in your project:
{
"mcpServers": {
"websearch-deepseek": {
"command": "npx",
"args": ["websearch-deepseek"],
"env": {
"DEEPSEEK_API_KEY": "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
"WEBSEARCH_MODEL": "deepseek-v4-flash",
"WEBSEARCH_THINKING": "enabled"
}
}
}
}Edit ~/.cursor/mcp.json:
{
"mcpServers": {
"websearch-deepseek": {
"command": "npx",
"args": ["websearch-deepseek"],
"env": {
"DEEPSEEK_API_KEY": "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
}
}
}
}Edit ~/.continue/config.json, add to mcpServers:
{
"mcpServers": [
{
"name": "websearch-deepseek",
"command": "npx",
"args": ["websearch-deepseek"],
"env": {
"DEEPSEEK_API_KEY": "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
}
}
]
}Edit ~/.pi/agent/settings.json, add to mcp:
{
"mcp": {
"websearch-deepseek": {
"command": "npx",
"args": ["websearch-deepseek"],
"env": {
"DEEPSEEK_API_KEY": "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
}
}
}
}Restart your AI coding assistant and ask a question that needs real-time information. The assistant will automatically call the web_search tool when needed.
Examples:
"What's new in React 19?"
"Search for Python 3.13 release date and major updates"
"What are the latest AI industry news?"
Variable | Required | Default | Description |
| ✅ Yes | — | DeepSeek API Key |
| ❌ No | — | Alternative API Key variable name |
| ❌ No |
| Model: |
| ❌ No |
| Thinking mode: |
| ❌ No |
| Max tokens for response |
Scenario | Recommended Model | Notes |
Daily search (default) |
| Fast, low cost, good quality |
Deep research |
| More detailed and accurate, slightly slower |
enabled (default): The model thinks before answering, producing higher quality results but consuming more tokens
disabled: Skips the thinking step for faster responses, suitable for simple queries
You can also use it directly from the terminal:
# Set API Key
export DEEPSEEK_API_KEY=sk-xxxxxxxxxxxxxxxx
# Search
npx websearch-deepseek search "Node.js LTS versions"
# JSON output
npx websearch-deepseek search --json "TypeScript 5.8"
# Specify model and disable thinking
npx websearch-deepseek search --model deepseek-v4-pro --no-thinking "Rust latest version"## Node.js Latest LTS Versions
As of July 2025, the latest Active LTS version of Node.js is Node.js 24.x...
| Version | Status | Initial Release | Support Until |
|---------|--------|-----------------|---------------|
| 24.x | Active LTS | 2025-05-06 | 2028-04 |
| 22.x | Maintenance | 2024-10 | 2027-04 |
### Recommendations
- New projects: Use Node.js 24 LTS
- Existing projects: Node.js 22 continues to receive security updates
---
### Sources (10):
1. [Node.js 24.0 is available...](https://...)
2. [Node.js — Node.js Releases](https://...)
...This tool uses the DeepSeek API, which charges per token. A single search typically consumes:
Component | Estimated Tokens |
Search + thinking | ~5,000–8,000 |
Generate answer | ~1,000–3,000 |
Total per search | ~8,000–15,000 tokens |
Check DeepSeek pricing for current rates.
MIT
搜索互联网获取当前、实时或事实性信息。 当你需要训练数据之外的信息时使用此工具—— 近期事件、当前数据、文档查询或事实核查。 返回基于完整网页内容生成的 AI 详细回答，附带来源 URL。 由 DeepSeek 原生联网搜索 API 驱动（服务端执行）。
English: Search the web for current, real-time, or factual information. Use this tool when you need information beyond your training cutoff — recent events, current data, documentation lookups, or fact-checking. Returns a detailed AI-generated answer based on full page content, plus source URLs. Powered by DeepSeek's native web search API.
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | 搜索关键词。请具体并包含相关关键词以获得更好结果。 / The search query. Be specific and include relevant keywords. | |
| explanation | No | 一句话解释为什么需要搜索，帮助理解上下文。 / One sentence explaining why this search is needed. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the burden. It discloses that the tool returns a 'detailed AI-generated answer based on full page content, plus source URLs,' and that it is 'powered by DeepSeek's native web search API (server-side execution).' This adequately sets expectations, though it omits potential limitations like result filtering or rate limits.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise in each language version, with the English text using four short sentences. Bilingual duplication doubles length, but it's necessary for multilingual support. Each sentence contributes meaningful information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple 2-parameter tool with no output schema, the description covers what it does, when to use it, what it returns, and its backend. This is fully sufficient for an agent to invoke it correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, and both parameters ('query' and 'explanation') have clear descriptions. The tool description itself adds no extra parameter-specific detail beyond the schema, so the baseline of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool performs web searches for current, real-time, or factual information. The verb 'search' and resource 'the web' are explicit, and it distinguishes itself from knowledge within the training cutoff, making its purpose unmistakable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It provides explicit 'when to use' guidance: 'when you need information beyond your training cutoff — recent events, current data, documentation lookups, or fact-checking.' This implies the inverse (don't use for well-known static facts) and clearly frames the tool's role.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.1web_searchScored across 1 tool
Only one tool exists, so there is no possibility of confusion between overlapping purposes. The tool's purpose is clear and unambiguous.
The single tool name 'web_search' follows a clear verb_noun pattern, which is consistent and descriptive. No naming conflicts or mixed conventions exist.
One tool is perfectly scoped for a web search server; additional tools would be redundant. The tool fully embodies the server's purpose without unnecessary bloat or thinness.
For a web search domain, the tool covers the complete expected functionality: fetching current, real-time, and factual information with sources. There are no obvious gaps or missing operations.
Docs: https://docs.keenable.ai/mcp-server Keenable is a free, remote MCP server that gives agents access to the web index. Search the web with ranked results and date/site filters, then fetch any indexed page as clean markdown. Works out of the box with no account or API key.
Live AI-native web search with citations. One tool for every MCP client. Flat per-request pricing.
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
