maru-deep-pro-search 0.9.0

Universal AI search MCP server — Perplexity-level quality with zero API keys. Multi-engine web scraping, intelligent ranking, and citation-native answers.

maru-deep-pro-search

Force your AI agent to research before it codes.
Zero API keys · Direct scraping · Citation-native answers

🇰🇷 한국어

🌐 Website · 📦 PyPI · 💻 GitHub

---

One-liner Install

macOS / Linux:

curl -sSL https://raw.githubusercontent.com/claudianus/maru-deep-pro-search/main/scripts/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/claudianus/maru-deep-pro-search/main/scripts/install.ps1 | iex

Or with pip:

pip install maru-deep-pro-search && maru-deep-pro-search setup

The setup wizard auto-detects your AI agent (Claude Code, Cursor, Kimi, Windsurf, etc.), backs up existing configs, injects MCP settings, and enforces research-first rules.

---

What it does

Your AI coding agent has a critical flaw: it answers from stale training data. maru-deep-pro-search fixes this by giving your agent live web search superpowers — and forcing it to use them first.

Capability	How
Search	Scrapes 7 engines directly. No API keys.
Rank	BM25 + authority/freshness/code-density scoring
Research	7-phase deep research pipeline with auto query expansion
Cite	Every result gets [1], [2] IDs — native citation architecture
Enforce	Setup CLI injects mandatory research-first rules into your agent

Core principle: 100% free, forever. No OpenAI, no Anthropic, no Google Search API, no SerpAPI.

---

Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│ User Query  │────▶│ Query Expand │────▶│ 7 Engines   │
└─────────────┘     └──────────────┘     └──────┬──────┘
                                                  │
                       ┌──────────────────────────┘
                       ▼
              ┌─────────────────┐
              │ BM25 + Metadata │
              │    Ranking      │
              └────────┬────────┘
                       │
         ┌─────────────┼─────────────┐
         ▼             ▼             ▼
    ┌─────────┐  ┌──────────┐  ┌──────────┐
    │trafilatura│  │htmldate │  │ 21-lang  │
    │+ content│   │+freshness│  │  code    │
    └────┬────┘  └────┬─────┘  └────┬─────┘
         └─────────────┼─────────────┘
                       ▼
              ┌─────────────────┐
              │ Cited Answer    │
              │ [1] [2] [3]     │
              └─────────────────┘

---

Before & After

	Before	After
Agent answers	From stale 2023 training data	From live web search
Sources	None, hallucinated	[1], [2] with real URLs
Setup	Manual MCP config per agent	One-liner auto-detects all agents
Cost	$5–50/mo API fees	$0 forever

---

8 Tools

Tool	Purpose
answer	Quick answer with inline citations
web_search	Scrape + rank + return cited results
search_with_citations	Pre-numbered sources for academic writing
fetch_page	Extract clean content from a single URL
fetch_bulk	Parallel fetch with deduplication
deep_research	7-phase pipeline: expand → search → rank → crawl → synthesize
stealthy_fetch	Anti-bot bypass for protected sites
parallel_search	Run multiple searches simultaneously

Decision tree:

• Quick answer? → answer
• Need sources? → web_search or search_with_citations
• Deep dive? → deep_research
• Blocked? → stealthy_fetch

---

Configuration

All environment variables are optional:

Variable	Default	Description
MARU_SEARCH_ENGINE	duckduckgo_lite	Default engine
MARU_SEARCH_MAX_RESULTS	10	Results per query
MARU_SEARCH_MAX_CONCURRENT	5	Parallel fetch limit
MARU_SEARCH_MAX_TOKENS_SOURCE	2500	Token budget per source
MARU_SEARCH_MAX_TOKENS_TOTAL	20000	Total output token budget
MARU_SEARCH_TIMEOUT	30.0	Fetch timeout (seconds)
MARU_SEARCH_RETRIES	3	Retry attempts

---

Testing

pytest tests/ -v

174 tests, all passing.

---

Contributing

PRs welcome. See CONTRIBUTING.md for coding style and PR guidelines.

See CHANGELOG.md for release history.

---

License

MIT © claudianus