/ Directory / Playground / DuckDuckGo MCP
● Community nickclyde ⚡ Instant

DuckDuckGo MCP

by nickclyde · nickclyde/duckduckgo-mcp-server

Web search and URL fetching — no API key, no signup. The simplest way to give an agent eyes on the open web.

duckduckgo-mcp-server uses DuckDuckGo's HTML endpoint for search results and adds a URL content extractor. No API key, no rate-limit dashboard, no plan tier — just install and use. Best for low-stakes lookups, getting links to authoritative pages, and quick fact-checking.

Why use it

Key features

Live Demo

What it looks like in practice

duckduckgo-mcp.replay ▶ ready
0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "duckduckgo-mcp": {
      "command": "uvx",
      "args": [
        "duckduckgo-mcp-server"
      ]
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "duckduckgo-mcp": {
      "command": "uvx",
      "args": [
        "duckduckgo-mcp-server"
      ]
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "duckduckgo-mcp": {
      "command": "uvx",
      "args": [
        "duckduckgo-mcp-server"
      ]
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "duckduckgo-mcp": {
      "command": "uvx",
      "args": [
        "duckduckgo-mcp-server"
      ]
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "duckduckgo-mcp",
      "command": "uvx",
      "args": [
        "duckduckgo-mcp-server"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json
{
  "context_servers": {
    "duckduckgo-mcp": {
      "command": {
        "path": "uvx",
        "args": [
          "duckduckgo-mcp-server"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add duckduckgo-mcp -- uvx duckduckgo-mcp-server

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use DuckDuckGo MCP

Fact-check a claim with web evidence

👤 Anyone wary of AI hallucinations ⏱ ~5 min beginner

When to use: You want Claude to verify a statistic or quote before quoting it.

Flow
  1. Search
    Use duckduckgo-mcp. Search for the original source of the claim 'Postgres 17 added unloggable replication slots'.✓ Copied
    → Result list with relevant URLs
  2. Read top results
    Fetch the top 2 official-looking results and quote what they actually say.✓ Copied
    → Quoted excerpts with URLs
  3. Verdict
    Is the claim accurate? If not, what's the closest true statement?✓ Copied
    → Verdict with citation

Outcome: Claims with sources, not vibes.

Pitfalls
  • Top result is a content-farm rephrasing of the actual doc — Prefer official domains (postgresql.org, github.com/<project>); skip listicles
Combine with: context7

Research three options for a tooling decision

👤 Engineers evaluating libraries ⏱ ~20 min beginner

When to use: You have to pick between A/B/C and want a quick comparative scan.

Flow
  1. Define criteria
    Criteria for comparing 3 ORMs: license, async support, last release, GitHub stars, known limitations.✓ Copied
    → Criteria locked
  2. Search per option
    For each ORM, search for: official site, latest release notes, recent critical issue.✓ Copied
    → Per-option link bundle
  3. Comparison table
    Build a comparison table from what the pages actually say. Cite sources.✓ Copied
    → Tabular comparison with citations

Outcome: Decision-ready comparison without 30 browser tabs.

Pitfalls
  • Stale info — last release date wrong — Always check official release page or GitHub releases, not third-party blogs
Combine with: git-mcp-idosal

Pull and summarize a specific URL

👤 Anyone reading articles via Claude ⏱ ~3 min beginner

When to use: You have a URL and want the gist + key quotes, not 1500 words of fluff.

Flow
  1. Fetch
    Use duckduckgo-mcp. Fetch <url> and return the cleaned content.✓ Copied
    → Readable text without nav/ads
  2. Summarize
    TL;DR in 5 bullets + 3 direct quotes worth keeping.✓ Copied
    → Bullets + quotes; no invented content

Outcome: 30-second comprehension of an article.

Pitfalls
  • Page is JavaScript-rendered — fetcher gets nothing — Fall back to a browser-MCP (browser-tools, chrome-devtools-mcp) for SPA pages

Combinations

Pair with other MCPs for X10 leverage

duckduckgo-mcp + context7

Two-tier doc lookup: Context7 for libraries, DuckDuckGo for everything else

Try context7 first; if it doesn't have the library, fall back to duckduckgo-mcp.✓ Copied
duckduckgo-mcp + git-mcp-idosal

Find a project on the web, then dive into its repo

Search 'fastest small embedding model 2026'. Pick top repo, switch to gitmcp for source.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
search query: str, max_results?, region? Open-ended web lookups Throttled by DDG; treat as ~30 req/min/IP
fetch_content url: str, max_length? After picking 1–3 promising URLs from search Throttled by target site

Cost & Limits

What this costs to run

API quota
DDG isn't an official API — informal limit ~30 req/min per IP; spike too high and you get blocked
Tokens per call
Search ~500–2000 tokens; fetch_content depends on page
Monetary
Free
Tip
Cache common lookups locally; don't search on every conversation turn

Security

Permissions, secrets, blast radius

Minimum scopes: Outbound HTTPS only
Credential storage: None — no key required
Data egress: duckduckgo.com + whatever sites you fetch

Troubleshooting

Common errors and fixes

All searches returning empty

DDG sometimes blocks rapid-fire requests — back off; verify with curl 'https://duckduckgo.com/?q=test'

Verify: Try a manual browser query
Fetch returns mostly nav junk

Some sites need a real browser; switch to a browser-MCP. Or set max_length lower to skip the tail

429 Too Many Requests

Wait 60s; reduce concurrency; the server has built-in throttle but bursty agent loops can still trip

Region-specific results wrong

Pass region parameter (us-en, jp-jp, etc.)

Alternatives

DuckDuckGo MCP vs others

AlternativeWhen to use it insteadTradeoff
Tavily MCPYou need higher-quality, agent-tuned search with reliable rate limitsRequires API key; paid above free tier
Exa MCPYou want neural search optimized for LLM consumersPaid; sign-up required
Firecrawl MCPCrawling/scraping rather than ad-hoc searchDifferent shape — site-wide jobs, not per-query

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills