Web Researcher MCP

1) Install Web Researcher MCP (fastest: uvx): Install uv (one-time), then run the MCP server via uvx. macOS/Linux: - curl -LsSf https://astral.sh/uv/install.sh | sh Then: - claude mcp add --scope user web-researcher -- uvx web-researcher-mcp This fetches the correct prebuilt binary for your OS and runs it (no Go toolchain needed). Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

2) Alternative installs (choose one): macOS (Homebrew): - brew install zoharbabin/tap/web-researcher-mcp - claude mcp add --scope user web-researcher -- web-researcher-mcp macOS/Linux (curl installer): - curl -fsSL https://raw.githubusercontent.com/zoharbabin/web-researcher-mcp/main/install.sh | sh Windows (PowerShell installer): - powershell -ExecutionPolicy Bypass -c "irm https://raw.githubusercontent.com/zoharbabin/web-researcher-mcp/main/install.ps1 | iex" Other options: - winget install zoharbabin.web-researcher-mcp - scoop bucket add zoharbabin https://github.com/zoharbabin/scoop-bucket && scoop install web-researcher-mcp - brew install --cask zoharbabin/tap/web-researcher-mcp - go install github.com/zoharbabin/web-researcher-mcp/cmd/web-researcher-mcp@latest - docker run -i --rm -e GOOGLE_CUSTOM_SEARCH_API_KEY=YOUR_KEY -e GOOGLE_CUSTOM_SEARCH_ID=YOUR_CX docker.io/zoharbabin/web-researcher-mcp:latest Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

3) Connect it to your MCP client (generic JSON config): If your app isn’t Claude Code’s CLI flow, add an MCP server entry to your client config. Example (Google PSE): { "mcpServers": { "web-researcher": { "command": "web-researcher-mcp", "env": { "SEARCH_PROVIDER": "google", "GOOGLE_CUSTOM_SEARCH_API_KEY": "YOUR_GOOGLE_API_KEY", "GOOGLE_CUSTOM_SEARCH_ID": "YOUR_SEARCH_ENGINE_ID" } } } } Example (Brave): { "mcpServers": { "web-researcher": { "command": "web-researcher-mcp", "env": { "SEARCH_PROVIDER": "brave", "BRAVE_API_KEY": "YOUR_BRAVE_API_KEY" } } } } Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

4) Run with zero config (DuckDuckGo fallback): If you don’t set SEARCH_PROVIDER or any API keys, Web Researcher MCP still works out of the box using DuckDuckGo as the built-in fallback (no API key required). This is useful to validate your setup before adding provider keys. Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

5) (Recommended) Add a search provider key to improve quality and unlock more features: Set SEARCH_PROVIDER and the matching key(s). Supported providers include DuckDuckGo (no key), Google PSE, Brave, Serper, SearchAPI.io, SearXNG, Tavily, Exa, and Hacker News. Example env vars: - SEARCH_PROVIDER=brave - BRAVE_API_KEY=... Or Google PSE: - SEARCH_PROVIDER=google - GOOGLE_CUSTOM_SEARCH_API_KEY=... - GOOGLE_CUSTOM_SEARCH_ID=... Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

6) (Optional) Configure multi-provider failover: To avoid downtime/rate limits, configure multiple providers so the server automatically falls back. Example: - SEARCH_ROUTING=brave,google,serper - BRAVE_API_KEY=... - GOOGLE_CUSTOM_SEARCH_API_KEY=... - GOOGLE_CUSTOM_SEARCH_ID=... - SERPER_API_KEY=... Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

7) Use “search lenses” to restrict results to trusted sources: When you want higher trust and less spam, use a built-in lens (e.g., medical, legal, academic, news) so searches are restricted to curated domains. Example prompt to your AI client: - “Search for recent findings on SGLT2 inhibitors using the clinical lens.” You can also create custom lenses by adding a JSON file under lenses/ with a domain allowlist. Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

8) Ask your AI to use the core tools (common workflows): Once connected, your AI can call tools such as: - web_search: search the web (optionally with lenses) - scrape_page: read a URL in full (web pages, PDFs, DOCX, PPTX, YouTube transcripts) - search_and_scrape: search then automatically read top results - news_search / image_search: when supported by your provider - academic_search / patent_search / legal_search / filing_search / econ_search / clinical_search: specialized databases - verify_citation / audit_bibliography / format_bibliography: citation workflows - sequential_search: multi-step deep research Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

9) Use guided prompt templates for repeatable research: Use the built-in prompt templates (where your client supports prompt selection) to run structured workflows: - comprehensive-research - fact-check - competitive-analysis - literature-review These templates guide multi-step research so you don’t have to specify every instruction. Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

10) Troubleshoot common issues (quick fixes): If tools fail with “API key” errors: set keys in your MCP client config env block (not only in shell profiles). If pages come back empty: JavaScript-heavy sites require Chromium; the binary auto-downloads Chromium on first use (~200MB). You can set CHROME_PATH to an existing Chrome install. Docker image includes Chromium. If cache is stale after upgrade: delete the cache directory (e.g., ~/Library/Caches/web-researcher-mcp/ on macOS) or set CACHE_DIR. If you hit provider rate limits (429): switch providers or configure SEARCH_ROUTING for failover. Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt

11) (Teams) Run in HTTP mode and connect clients to a shared endpoint: For shared/team setups, run the server as an HTTP service and connect MCP clients to http://localhost:3000/mcp/. Example: - PORT=3000 OAUTH_ISSUER_URL=https://auth.example.com OAUTH_AUDIENCE=https://api.example.com ./web-researcher-mcp Docker Compose example: services: web-researcher: image: zoharbabin/web-researcher-mcp ports: - "3000:3000" environment: PORT: "3000" SEARCH_PROVIDER: brave BRAVE_API_KEY: ${BRAVE_API_KEY} Source: https://github.com/zoharbabin/web-researcher-mcp?ref=producthunt