SourceWeave Web Search
Run privacy-minded web research from Claude or a CLI using your own SearXNG and Crawl4AI stack instead of a closed SaaS search API.
Overview
SourceWeave Web Search is a MCP server for the Idea phase that performs local SearXNG-backed web research and Crawl4AI crawling for coding agents.
What is this MCP server?
- Fully local MCP server and CLI for web research workflows
- SearXNG query template via SOURCEWEAVE_SEARCH_SEARXNG_BASE_URL with required <query> placeholder
- Crawl4AI integration through SOURCEWEAVE_SEARCH_CRAWL4AI_BASE_URL for page retrieval
- Optional Redis or Valkey cache via SOURCEWEAVE_SEARCH_CACHE_REDIS_URL
- uvx runtime hint on PyPI package sourceweave-web-search v0.5.0
- Package version 0.5.0 on PyPI identifier sourceweave-web-search
- Three documented environment variables for SearXNG, Crawl4AI, and optional Redis cache
- Dual surface: MCP stdio server and CLI per project description
Community signal: 2 GitHub stars.
What problem does it solve?
Closed search APIs and manual tab switching slow down early validation when you want agent-led research on infrastructure you control.
Who is it for?
Privacy-conscious indie builders who already run or will run SearXNG and Crawl4AI and want MCP-native research in idea discovery.
Skip if: Builders who want zero-infra Google results in one API key or who are only optimizing published site SEO in the launch phase.
What do I get? / Deliverables
After configuring URLs, your agent or CLI can search and crawl the web locally with optional cache acceleration for repeat prompts.
- Agent-callable search and crawl results from your stack
- CLI-accessible web research for scripted validation workflows
- Cached repeat queries when Redis is configured
Recommended MCP Servers
Journey fit
Idea-phase research is when solo builders validate markets, competitors, and technical feasibility before committing to build. Research subphase matches a fully local MCP plus CLI that searches and crawls the open web under your infrastructure.
How it compares
Self-hosted research MCP, not a hosted Perplexity clone or Skillselion marketplace skill.
Common Questions / FAQ
Who is SourceWeave Web Search for?
Solo developers and agents-users who self-host search and crawl services and need MCP plus CLI access for early-stage web research.
When should I use SourceWeave Web Search?
Use it during idea and competitor research before you build, when local SearXNG and Crawl4AI are acceptable setup costs for trustworthy grounding.
How do I add SourceWeave Web Search to my agent?
Install sourceweave-web-search from PyPI (uvx hint), set SearXNG and Crawl4AI environment variables, optionally set Redis cache URL, and register the stdio MCP server in your client.