setup

# Web Search Services Setup Verification Verify which search and scraping services are connected and operational. ## Service Availability Check Run one lightweight call per service to confirm connectivity. Not all services need to be active — the plugin works with any subset. ### 1. Firecrawl ``` Tool: firecrawl_search Input: { "query": "test", "limit": 1, "sources": ["web"] } ``` **Expected:** Returns search results. If error, check `FIRECRAWL_API_TOKEN` — the env var this plugin's `.mcp.json` maps to the server's `FIRECRAWL_API_KEY`. Without a key only `firecrawl_scrape`, `firecrawl_search` and `firecrawl_parse` work (rate-limited). Optional: `firecrawl_credit_usage` confirms the key and shows the balance. The server needs Node.js 22+. ### 2. Exa ``` Tool: web_search_exa Input: { "query": "test", "numResults": 1 } ``` **Expected:** Returns results with URLs. If error, check `EXA_API_KEY`. `agent_run` and `web_search_advanced_exa` are opt-in on the bundled server (`ENABLED_TOOLS` environment variable) — their absence is not an error. ### 3. Perplexity ``` Tool: perplexity_search Input: { "query": "test", "max_results": 1, "search_type": "fast" } ``` **Expected:** Returns a ranked result (title, URL, snippet). If error, check `PERPLEXITY_API_KEY`. The answer tools take `messages`, not `query`: ``` Tool: perplexity_ask Input: { "messages": [{ "role": "user", "content": "What is 2+2?" }] } ``` ### 4. Jina ``` Tool: read_url Input: { "url": "https://example.com", "question": "What is this page about?" } ``` **Expected:** Returns the passage(s) answering the question (omit `question` for the full page). `read_url` works without an API key (rate-limited); search tools need a key. The server exposes 12 tools. ### 5. Context7 ``` Tool: resolve-library-id Input: { "query": "hooks", "libraryName": "React" } ``` **Expected:** Returns library ID like `/facebook/react`. Both `query` and `libraryName` are required. API key optional — keyless works with low rate limits; a key (context7.com/dashboard) grants higher limits and private repos. ### 6. Media Services (Pexels / Unsplash — REST) These are plain REST APIs, not MCP servers. Check the keys with one request each (the output shows only status codes, never the keys): ```bash curl -s -o /dev/null -w "pexels %{http_code}\n" -H "Authorization: ${PEXELS_API_KEY}" \ "https://api.pexels.com/v1/search?query=nature&per_page=1" curl -s -o /dev/null -w "unsplash %{http_code}\n" -H "Authorization: Client-ID ${UNSPLASH_ACCESS_KEY}" \ "https://api.unsplash.com/search/photos?query=nature&per_page=1" ``` **Expected:** `200` for each configured service; `401` means a missing or wrong key. Both keys are optional (`PEXELS_API_KEY`, `UNSPLASH_ACCESS_KEY` — set them in the environment that launches Claude Code). ## Service Status Summary After running checks, report which services are available: | Service | Status | Capabilities | |---------|--------|-------------| | Firecrawl | Connected / Not available | scrape (incl. JSON format), crawl, search, parse, agent, interact, monitors, research/developer search, Alexandria | | Exa | Connected / Not available | semantic search, page fetch (+ advanced search, `agent_run` when enabled) | | Perplexity | Connected / Not available | search, AI answers, research, reasoning | | Jina | Connected / Not available | read pages (batch + question mode), search, images, rerank, deduplicate | | Context7 | Connected / Not available | framework documentation search | | Pexels (REST) | Key valid / No key | stock photos and videos | | Unsplash (REST) | Key valid / No key | stock photos | ## Working with Partial Availability Not all services need to be active — the plugin works with any subset. Here's what you can do with common combinations: | Available Services | You Can Do | |-------------------|------------| | Firecrawl only | Scrape, crawl, search, structured JSON extraction, agent research, live-page interaction, parse files, monitor changes | | Jina only | Read pages (batch + question mode), search web/papers/images, rerank, deduplicate | | Exa only | Semantic search, page fetch, domain-scoped search | | Perplexity only | AI Q&A, research, reasoning with citations | | Context7 only | Framework/library documentation search | | Firecrawl + Jina | Full scraping pipeline with fallbacks | | Any service + Context7 | Dev workflow with doc search | If a recommended tool is unavailable, the mcp-patterns skill routing table shows fallback alternatives for every task. ## Common Issues | Symptom | Cause | Fix | |---------|-------|-----| | `npx` command not found | Node.js not installed | Install Node.js 22+ (required by `firecrawl-mcp`; the other servers need 20+) | | 401 Unauthorized | Invalid API key | Check key in plugin settings or regenerate | | Connection timeout | Network or firewall | Check internet connectivity | | MCP server not starting | npm package issue | Run `npx -y firecrawl-mcp` manually to see errors | | Tool not found | MCP server not configured | Check `.mcp.json` or Claude Code MCP settings | ## What This Skill Does NOT Cover - Creating API accounts — visit each service's website to sign up - Configuring MCP servers from scratch — the plugin's `.mcp.json` handles this automatically