mcp-local-rag

Local web search for your LLM over MCP. No API keys, and results are ranked on your machine. Formerly mcp-local-rag.

%%{init: {'theme': 'base'}}%%
flowchart TD
    A[LLM] -->|query| S["Search with ddgs (auto: several engines, no keys)"]
    S -->|"rag_search, deep_research"| R["Re-rank results locally (MobileBERT on numpy)"]
    R --> P[Fetch top pages, keep the main content]
    P -->|page text + URLs| A
    S -->|search| T1["Tier 1: pick pages by their snippets"]
    T1 --> T2["Tier 2: split into passages, keyword match (BM25)"]
    T2 -->|few hits| OUT["Best passages + URLs (about 800 tokens)"]
    T2 -->|too many or none| T3[Tier 3: rank them by meaning]
    T3 --> OUT
    OUT --> A

What's in it:

• No API keys. Searches go through ddgs. Its auto mode asks several engines at once and skips blocked ones.
• Ranked locally. A small MobileBERT model on plain numpy re-ranks results, so only the best pages get fetched.
• Clean text. Pages are fetched with browser impersonation, and only the main content is kept: no menus, cookie banners or footers. Code blocks and tables keep their lines. The same page under two URLs comes back once.
• Passages instead of pages. search returns just the paragraphs that answer the question, about a tenth of the tokens. When nothing matches well, it can ask you whether you want the full pages instead.
• Says what it left out. Results list skipped pages with a reason (blocked, PDF, duplicate...), cut pages are marked truncated, and fallbacks and engine failures show up as warnings. Google blocks often, so it falls back to auto on its own.
• Brings its own guide. It serves a skill and a research prompt, so clients pick up how to use it without setup.

Tool	Use it for
search	Start here: returns only the relevant passages. Pass keywords for better hits
rag_search_ddgs	Full pages as text (the original mode)
rag_search_google	Same, via Google (falls back to auto when blocked)
deep_research	Several search terms across backends, merged and de-duplicated
deep_research_ddgs / deep_research_google	deep_research shortcuts

Backends: auto (default), brave, duckduckgo, google, grokipedia, mojeek, startpage, wikipedia, yahoo. auto and brave are the dependable ones.

Install

With uv:

{
  "mcpServers": {
    "mcpsurfr": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/nkapila6/mcpsurfr", "mcpsurfr"]
    }
  }
}

Or with Docker (amd64 and arm64):

{
  "mcpServers": {
    "mcpsurfr": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "--init", "ghcr.io/nkapila6/mcpsurfr:latest"]
    }
  }
}

Full-page tools share a 40k-character budget per call (about 10k tokens). To change it, add "env": {"MCPSURFR_MAX_CHARS": "80000"} to the server entry.

Not sure where your config lives? See the MCP quickstart. Tested with Claude Desktop, Claude Code, Cursor and Goose.

Coming from mcp-local-rag? Your old config keeps working, since the old URL redirects and the mcp-local-rag command still exists. Switching to the config above gets you the new name and drops the --python=3.10 pin (3.10 reaches end of life this month).

Skill and prompt

The mcpsurfr-search skill tells the model which tool and backend to pick and how to cite. The server serves it as skill://mcpsurfr-search/SKILL.md, so most clients need nothing extra. To add it to Claude Desktop by hand: Settings → Skills → Add from folder, then pick src/mcpsurfr/skills/mcpsurfr-search/.

The research(topic) prompt plans a few search terms, runs deep_research and writes a cited summary.

Example

Asking Claude Desktop about something newer than its training data:

---

Security audit by MseeP · Issues and PRs welcome · MIT licensed · Buy me a coffee if it helps you