zero-api-key-web-search

RAG & Memory 💻 Python ⚖️ MIT 🟡 Quiet lately
16 stars

Zero-API-Key Web Search is a local-first, MCP-native search and evidence-verification toolkit for AI agents. It provides live web search, LLM-optimized context extraction, claim verification with weighted evidence scoring, and citation-ready evidence reports without requiring an API key by default. It is designed for developers building AI agents that need to ground answers in current web data.

✨ Key features

  • Free by default with DuckDuckGo, no API key required
  • MCP server with 8 tools for Claude Code, Cursor, Copilot
  • Claim verification with evidence scoring and verdicts
  • Multi-engine SERP via Bright Data (7 engines, geo-targeting)
  • Web Unlocker for blocked or geo-restricted pages
  • Built-in goggles presets for reranking search results

🎯 Use cases

  • Ground AI agent answers with live web search and citations
  • Verify claims and generate evidence reports for fact-checking
  • Access blocked or geo-restricted pages for research
  • Cross-validate search results using multiple providers
  • Serve search tools to MCP-compatible agents

📦 Installation

🧰 Requirements: Python 3.x, pip install; optional Bright Data API key for production features, optional self-hosted SearXNG instance.

pip install zero-api-key-web-search

🚀 Usage

Search the web — no API key needed
zero-search "Python 3.13 release" --json
Build citation-ready LLM context
zero-context "Python 3.13 stable release" --goggles docs-first
Read a page
zero-browse "https://docs.python.org/3/whatsnew/" --json
Verify a claim
zero-verify "Python 3.13 is the latest stable release" --deep --json
Full evidence report
zero-report "Python 3.13 stable release" \
  --claim "Python 3.13 is the latest stable release" --deep --json

⚠️ Good to know

The verification model is a heuristic evidence classifier, not a proof engine; it does not perform fact-level proof or logical entailment.

❓ FAQ

Do I need an API key to use this?

No, the default provider (DuckDuckGo) works without any API key or account. API keys are only needed for optional production features like Bright Data.

What is the MCP server and how do I use it?

The MCP server exposes 8 tools (e.g., search_web, verify_claim) and can be configured in MCP-compatible clients like Claude Code, Cursor, and Copilot using the command zero-mcp.

How does claim verification work?

It uses the evidence-aware heuristic v3 model, which searches for the claim, scores sources on keyword overlap, source quality, and freshness, and classifies them as supporting, conflicting, or neutral to produce a verdict.

Can I access blocked or geo-restricted pages?

Yes, if you configure the Bright Data Web Unlocker zone, the browse_page tool automatically retries blocked pages (403/429) through the unlocker, or you can force it with --use-unlocker always.

📊 Repository

🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.