Web Search Plus

• You want your AI agent to search the web without hardcoding a single provider — auto-routing means shopping queries go to Google results, research questions go to deep-research engines, and privacy-sensitive queries go to SearXNG automatically.
• You need multilingual or international search coverage (e.g. news in German, policy updates across languages) that a single English-first provider can't handle well.
• You're building a RAG pipeline and need real-time, citation-rich, or semantically indexed results depending on the query type.
• You want to start free — SearXNG is self-hosted at $0, and Serper, Tavily, and Exa all offer free monthly tiers totaling 4,500+ searches.

• You only ever need one specific provider and have no interest in auto-routing or fallback — a simpler single-provider skill would have less overhead.
• You require SearXNG on a private/internal network without setting SEARXNG_ALLOW_PRIVATE=1, as SSRF protections will block private IP ranges by default.
• You need all 7 providers active simultaneously — at least one API key is required, and providers without keys are simply skipped rather than returning an error.

The skill analyzes each query using multi-signal detection (shopping keywords, research intent, URL patterns, recency signals, privacy keywords) and routes it to the most appropriate provider. Every response includes a routing object with the chosen provider, confidence score, and whether a fallback was used.

Access Serper (Google results), Tavily (deep research), Querit (multilingual AI search), Exa (neural/semantic similarity), Perplexity (AI-synthesized answers with citations via Kilo Gateway), You.com (RAG-optimized real-time), and SearXNG (self-hosted privacy search) — all through a single unified script and output format.

All provider keys are optional — you only need at least one configured provider to start. Add more keys incrementally to expand routing coverage. SearXNG requires no API key at all, just a self-hosted instance URL.

If a provider hits a rate limit, times out, or returns an error, the skill automatically retries with the next available configured provider. The response flags routing.fallback_used: true so you always know what happened.

Force a specific provider with -p provider when auto-routing isn't the right fit for a query. Use --explain-routing to see exactly why the skill chose a given provider, making it easy to tune or override behavior.

The SearXNG instance URL is validated with defense-in-depth: only http/https schemes are allowed, cloud metadata endpoints are blocked, and hostnames are resolved to catch private/RFC1918 IPs. Operators self-hosting on private networks can opt in with SEARXNG_ALLOW_PRIVATE=1.

Queries like "iPhone 16 Pro Max price" or "best pizza near me" are automatically routed to Serper, which returns Google Shopping and local results. No configuration needed — the routing engine detects price and location intent.

Research questions like "how does HTTPS encryption work" or "transformer architecture papers" are routed to Tavily or Exa respectively, retrieving full-page content, academic sources, and semantically relevant results rather than surface-level snippets.

Agents building retrieval-augmented generation workflows can route queries to You.com for RAG-optimized results or Perplexity for AI-synthesized answers with citations, ensuring up-to-date grounding for language model responses.

Deploy a self-hosted SearXNG instance and route queries through it for zero-tracking, no-profiling search. The skill validates the instance URL for SSRF safety and supports multi-engine aggregation (e.g. --engines "google,bing").