api-reference

# Web Search Services API Reference Curated REST API endpoints for the 4 core services plus the Pexels and Unsplash media APIs. For full documentation, visit each service's official docs. ## Authentication Summary | Service | Auth Header | Key Format | Docs | |---------|------------|------------|------| | Firecrawl | `Authorization: Bearer KEY` | `fc-xxx` | https://docs.firecrawl.dev | | Exa | `x-api-key: KEY` (also accepts `Authorization: Bearer KEY`) | UUID string | https://exa.ai/docs | | Perplexity | `Authorization: Bearer KEY` | String | https://docs.perplexity.ai | | Jina | `Authorization: Bearer KEY` | `jina_xxx` | https://jina.ai/reader | | Pexels | `Authorization: KEY` (no `Bearer`) | String | https://www.pexels.com/api/documentation/ | | Unsplash | `Authorization: Client-ID ACCESS_KEY` | Access key | https://unsplash.com/documentation | ## Service Endpoints ### Firecrawl — `https://api.firecrawl.dev` | Method | Path | Description | |--------|------|-------------| | POST | `/v2/scrape` | Scrape single URL | | POST | `/v2/search` | Web search with content | | POST | `/v2/crawl` | Start site crawl | | GET | `/v2/crawl/{id}` | Check crawl status | | POST | `/v2/crawl/params-preview` | Preview crawler options from a natural-language prompt | | POST | `/v2/map` | Map site URLs | | POST | `/v2/extract` | Legacy LLM structured extraction (maintenance mode — prefer `/v2/scrape` with a `json` format, or `/v2/agent`) | | POST | `/v2/batch/scrape` | Batch scrape multiple URLs | | POST | `/v2/agent` | Autonomous research agent (default model `spark-2`; `GET /v2/agent/{jobId}` for status) | | POST | `/v2/scrape/{jobId}/interact` | Run a prompt or code in the live browser session of a scrape (`DELETE` the same path to stop) | | POST | `/v2/interact` | Create a standalone interact session (`/v2/interact/{sessionId}/execute` to run code, `DELETE /v2/interact/{sessionId}` to stop) | | POST | `/v2/parse` | Parse files (PDF, DOCX, …) into LLM-ready output | | POST / GET | `/v2/monitor` | Create / list page-change monitors (CRUD + run + checks) | | GET / POST | `/v2/search/developer` | Search the developer index (GitHub issues, PRs, READMEs, docs) | | GET | `/v2/search/research/papers` | Search academic papers (`/{id}` for one paper, `/{id}/similar` for related) | | GET | `/v2/team/credit-usage` | Remaining credits (`/v2/team/credit-usage/historical` for history) | See `references/firecrawl-api.md` for curl examples. ### Exa — `https://api.exa.ai` | Method | Path | Description | |--------|------|-------------| | POST | `/search` | Semantic web search (single powerful endpoint) | | POST | `/contents` | Get text/highlights/summary for URLs or result ids | | POST | `/answer` | Direct answer with citations | | POST | `/agent/runs` | Start an Exa Agent run (multi-step research) | | GET | `/agent/runs/{id}` | Poll an agent run | | POST | `/monitors` | Create a monitor (`GET /monitors`, `/monitors/{id}`, `/monitors/{id}/trigger`, `/monitors/{id}/runs`) | | POST | `/batches` | Batch API for bulk requests (`GET /batches/{id}`) | Exa also offers a Websets API — see https://exa.ai/docs for details. The URL-similarity endpoint is deprecated; describe the source page in a normal `/search` query instead. See `references/exa-api.md` for curl examples. ### Perplexity — `https://api.perplexity.ai` | Method | Path | Description | |--------|------|-------------| | POST | `/v1/agent` | Agent API (presets or models + web search) — the primary path | | POST | `/search` | Web search (no `/v1` prefix) | | POST | `/v1/sonar` | Legacy Sonar chat completion (support ended 2026-09-27; requests are being rerouted to the Agent API — migrate) | | POST | `/v1/embeddings` | Text embeddings (`pplx-embed-v1-0.6b`/`4b`; contextualized variants via the contextualized embeddings endpoint) | | POST | `/router/v1/chat/completions` | Gateway API (OpenAI-compatible) | See `references/perplexity-api.md` for curl examples. ### Jina — Multiple base URLs | Method | URL Pattern | Description | |--------|------------|-------------| | GET | `https://r.jina.ai/{URL}` | Read any URL as markdown | | GET | `https://s.jina.ai/?q={QUERY}` | Web search | | POST | `https://api.jina.ai/v1/embeddings` | Text embeddings | | POST | `https://api.jina.ai/v1/rerank` | Rerank documents | | POST | `https://api.jina.ai/v1/classify` | Text classification | See `references/jina-api.md` for curl examples. ### Pexels and Unsplash — media search | Method | URL | Description | |--------|-----|-------------| | GET | `https://api.pexels.com/v1/search` | Search photos (`query`, `orientation`, `size`, `color`, `locale`, `per_page`) | | GET | `https://api.pexels.com/v1/videos/search` | Search videos (videos live under `/v1/videos/`) | | GET | `https://api.pexels.com/v1/videos/popular` | Popular videos (`min_width`, `min_height`, `min_duration`, `max_duration`) | | GET | `https://api.unsplash.com/search/photos` | Search photos | | GET | `<photo.links.download_location>` | Unsplash download event — required when a photo is used | Rate limits: Pexels 200 req/hour and 20,000 req/month; Unsplash 50 req/hour (demo) or 1,000 req/hour (production). Unsplash API rules (hotlink `photo.urls.*`, download event, attribution) are in `mcp-patterns/references/media-tools.md`.