Searxng Search
Run local SearXNG metasearch from your agent to find packages, docs, and web results in JSON without relying on a single search API.
Overview
SearXNG Search is an agent skill most often used in Idea (also Build integrations, Validate scope) that runs local SearXNG metasearch with category filters and JSON output for package and web discovery.
Install
npx skills add https://github.com/ypares/agent-skills --skill searxng-searchWhat is this skill?
- Portable SearXNG via Podman or Docker with start/stop helper scripts
- Category-based filtering (IT, cargo, packages, scientific, repos) plus JSON for agents
- Documented status for 14 package repositories in references/package-engine-status.md
- PyPI workarounds via direct API and qypi CLI when metasearch gaps appear
- Nushell searx helper and curl+jq patterns for scripted queries
- 14 package repositories documented in package-engine-status
- category guide covering IT, repos, scientific, and package engines
Adoption & trust: 661 installs on skills.sh; 27 GitHub stars; 2/3 security scanners passed (skills.sh audits).
What problem does it solve?
You need federated search across repos and the web from an agent session without locking into one commercial search API or brittle scrape chains.
Who is it for?
Solo builders who want offline-friendly, scriptable search during research and dependency picking on Linux/macOS with Podman or Docker.
Skip if: Teams that forbid local containers, need managed enterprise search compliance, or only want a hosted SaaS search with no self-host ops.
When should I use this skill?
You need programmatic web or multi-repo package search on localhost:8888 with JSON output during agent sessions.
What do I get? / Deliverables
You get a running local SearXNG endpoint, repeatable query patterns, and JSON results you can paste into specs, dependency lists, or competitor notes.
- Running SearXNG container
- JSON search results for agent consumption
- Repeatable searx or curl query snippets
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Canonical shelf is Idea/research because the skill’s primary payoff is discovering options, repos, and references before you commit to a stack or scope. Research subphase fits unified category-filtered search across IT, scientific, and 14 package repositories with programmatic JSON output.
Where it fits
Compare async Rust crates by running cargo-category queries before picking Tokio versus alternatives.
Validate whether an npm package ecosystem has maintained forks using packages-category limits.
Pull Docker Hub image tags and docs links while wiring a sidecar in compose.
How it compares
Use as a self-hosted metasearch integration—not as a drop-in MCP server or a single-site npm search bookmark.
Common Questions / FAQ
Who is searxng-search for?
Indie and solo developers using Claude Code, Cursor, or similar agents who self-host tooling and want JSON search across web and package categories during research and build decisions.
When should I use searxng-search?
During Idea research to compare libraries and repos, during Validate when scoping stack choices, and during Build integrations when you need category-filtered package lookup from the agent.
Is searxng-search safe to install?
It runs containers and network calls on your machine; review the Security Audits panel on this Prism page and inspect scripts/start-searxng and image sources before enabling in CI or production hosts.
SKILL.md
READMESKILL.md - Searxng Search
# SearXNG Search Skill Enhanced web and package repository search capabilities using SearXNG metasearch engine. **Fully portable** - works on any machine with podman or docker, no external dependencies. ## What This Provides - **Unified search interface** across multiple package repositories (npm, Cargo, Docker Hub, etc.) - **Category-based filtering** for targeted searches (IT, repos, scientific publications, etc.) - **JSON output** for programmatic consumption - **Workarounds for PyPI** (direct API + qypi CLI tool) - **Nushell helper script** for convenient command-line usage ## Quick Start **1. Start SearXNG:** ```bash start-searxng --detach ``` **2. Use the search helper:** ```bash searx "tokio" --category cargo searx "express" --category packages searx "rust async" --category it --limit 5 ``` **3. Or use curl directly:** ```bash curl -s "http://localhost:8888/search?q=serde&format=json&categories=cargo" | jq '.results[0:3]' ``` **4. Stop when done:** ```bash podman stop searxng # or: docker stop searxng ``` ## Files - **SKILL.md**: Main documentation with quick reference and common patterns - **references/package-engine-status.md**: Test results for all 14 package repositories ⭐ - **references/category-guide.md**: Comprehensive guide to all search categories - **references/pypi-direct-search.md**: PyPI workarounds (API + qypi CLI) - **references/agent-usage.md**: Guide for AI agents using SearXNG - **scripts/start-searxng**: Bash script to start SearXNG container (portable!) - **scripts/searx**: Nushell helper script with colored output ## What Works ✅ **13/14 package repositories working**, including: - **Haskell (Hoogle/Hackage)** - packages & functions ⭐ - **JavaScript (npm)** - **Rust (crates.io, lib.rs)** - **Ruby (RubyGems)** - **PHP (Packagist)** - **Erlang/Elixir (Hex)** - **Perl (MetaCPAN)** - **Dart/Flutter (pub.dev)** - **Go (pkg.go.dev)** - **Docker Hub** - **Alpine/Void Linux packages** ✅ **General web search**: Multiple engines (DuckDuckGo, Startpage, etc.) ✅ **GitHub/GitLab**: Repository and code search ✅ **Academic papers**: arXiv, PubMed, Google Scholar, etc. ❌ **PyPI (Python)**: Broken due to bot protection - use direct API or qypi CLI instead **See `references/package-engine-status.md` for detailed test results!** ## Requirements - **podman** or **docker** (auto-detected) - **curl** (for API access) - **jq** (optional, for JSON parsing) - **nushell** (optional, for the `searx` helper script) No installation needed - runs in a container! ## Next Steps 1. Try the helper script for common searches 2. Explore different categories (see category-guide.md) 3. For PyPI searches, use `uvx qypi search <term>` or direct API ## Tips - Use `categories=packages` for multi-repo package search - Use `categories=cargo` specifically for Rust crates - Combine categories: `categories=packages,it,repos` - Filter results by engine in jq: `select(.engines[] == "npm")` # SearXNG for AI Agents This guide explains how AI agents (like Claude) can use SearXNG for enhanced search capabilities. ## Why Use SearXNG? **Problem**: Built-in WebSearch tools often have limitations: - Limited to ~10 results per query - Can't search specific package repositories (PyPI, npm, cargo) - No direct control over search engines used - May have rate limits or restrictions **Solution**: SearXNG provides: - Direct access to 100+ search engines - Specialized package repository search - Full control over categories and filters - Unlimited local queries - JSON API for programmatic access ## Agent Workflow ### 1. Start SearXNG (if not running) ```bash # Check if running curl -sf http://localhost:8888/ > /dev/null || \ start-searxng --detach ``` **When to start:** - At the beginning of a search-intensive task - When user explicitly requests package searches - When WebSearch tool returns insufficient results ### 2. Choose the Right Category | User Request | Category | Example | |--------------|----------|---------| | "Fi