# AnankeSearch API > Independent search for people and AI agents: search the web, news, videos, images, code and what people say, with filters; read any public page as Markdown or JSON; act in a real browser; watch pages for changes. REST API and MCP server. ## Basics - Base URL: the domain serving this file (for example https://anankesearch.com) - Auth: header `Authorization: Bearer as_live_` (free account: 500 credits, no card) - Without a key: a small public sandbox for trying the API - Format: JSON over HTTPS. Errors: `{"detail": {"error": "..."}}` with 401, 402 (credits), 403 (non-public URL), 422, 429 - Credits: search 1 (AI answer 2), scrape 1, map 1, crawl 1 per page, extract 2, interact 2, monitor 1, index 1, feedback free - Full documentation: /documentation · Playground: /playground · OpenAPI: /openapi.json ## POST /v1/search Request: ```json {"query": "elephant", "type": "videos", "time_range": "month", "language": "en", "site": null, "max_results": 10, "mode": "raw", "include_content": false, "social": "auto", "safe": true} ``` - type: web | news | videos | images | code | social. A query that names a medium ("elephant video") switches type automatically (`type_auto: true`); `strict_type: true` prevents it. - time_range: day | week | month | year. language: fr, en, es, pt, de... site: only that site. - mode "answer": adds an AI answer with citations. include_content: adds each page's Markdown. - social: people's voices from forums, open networks and public posts; `echo` tells how many independent voices agree (corroborated, discussed, amplified, single). - context: a few words about the conversation (ranking only). user: your end user's id (private index). Response: `results[]` with title, url, snippet, kind, score, date, and per type: videos (platform, thumbnail, duration, author, published, embed), images (image, thumbnail, width, height, license, source), news (source, published, thumbnail), code (language, stars). Plus answer_direct, answer, social[], echo[], notice, credits_used, remaining_credits, latency_ms. ## Read the web - POST /v1/scrape `{"url", "max_chars"}` → title, content (Markdown). When a page is down or gone, its latest archived copy is returned with `archived: true`, `archived_at`, `archive_url`. - POST /v1/map `{"url", "limit"}` → domain, total_links, links[] - POST /v1/crawl `{"url", "limit" (1-25), "max_depth" (1-4)}` → base_url, pages_crawled, data[] (url, title, content) - POST /v1/extract `{"url", "prompt", "schema"}` → data (JSON) - POST /v1/interact `{"url", "actions": [{"type": "click|type|scroll|wait", "selector", "text", "direction", "seconds"}], "wait_ms"}` → url, content - POST /v1/monitor `{"url"}` → has_changed, change_count, last_checked ## Personal index - POST /v1/feedback `{"url", "query", "user"}`: your user opened this result (free) - POST /v1/index `{"url", "title", "text", "user"}`: add your own document - DELETE /v1/index `{"user"}`: forget everything learned ## MCP - Endpoint: POST /v1/mcp (JSON-RPC 2.0: initialize, tools/list, tools/call), same Bearer key - Tools: anankesearch_web_search, anankesearch_scrape_url, anankesearch_map_domain, anankesearch_crawl_site, anankesearch_extract_data, anankesearch_interact_page, anankesearch_monitor_page