API changelog

This page tracks the public API at api.serpkite.com: endpoints, parameters, response fields, errors and billing rules. For changes to Google’s results pages themselves, see the Google SERP changelog.

Compatibility policy. New endpoints, optional parameters and response fields are additive and can appear at any time, so ignore fields you don’t recognize. Removing or renaming anything, or changing what a call costs, is a breaking change: it will get advance notice here and a new API version.

2026-10-02: AI Overview and AI Mode removed

Removed

• The ai_overview request parameter and response field on /v1/search (including format=compact and format=markdown output). Sending ai_overview now returns 400 invalid_request (“AI Overviews are no longer returned; remove it”).
• The POST /v1/ai-mode endpoint.
• The ai_overview tool on the remote MCP server.

Everything else on /v1/search (organic results, answer box, knowledge graph, People Also Ask, related searches, top stories, local pack and ads) and every other endpoint is unchanged, as are the credit prices.

v1.0.0 (2026-09-29): the v1 API

SerpKite’s own versioned API, published as an OpenAPI 3.1 spec. It replaces the unversioned preview routes.

Breaking changes from the preview

• Paths: every route is under /v1: /v1/search, /v1/news, /v1/images, /v1/videos, /v1/maps, /v1/places, /v1/reviews, /v1/shopping, /v1/scholar, /v1/patents, /v1/autocomplete, /v1/lens, /v1/ai-mode, /v1/webpage, /v1/rank, /v1/account, /v1/status, /v1/batches. GET /customsearch/v1 is unchanged.
• Auth: Authorization: Bearer skt_live_… (or ?api_key= on GET). The legacy key header is no longer accepted.
• Parameters: country and language set the market and interface language. Reviews take place_id (or cid/fid), sort and page_token. The endpoint is chosen by the path, not a body field.
• Strict validation: unknown parameters return 400 invalid_request, and the message names the replacement (for example unknown parameter "type": use the endpoint path).
• Response envelope: every vertical returns request (the normalised request), results (its primary list: organic rows, images, news, places, reviews, papers, suggestions…), vertical extras and meta. All keys are snake_case (ai_overview, knowledge_graph, people_also_ask, related_searches, displayed_link, image_url…). fields paths follow suit: results.title,results.link,ai_overview.
• Batches: POST /v1/batches with {"endpoint", "requests": [1–100 bodies], "webhook_url"} queues one half-price job per request. Realtime endpoints take a single JSON object; array bodies and the batch mode flag are gone.
• MCP: the remote server is at https://api.serpkite.com/v1/mcp, with Bearer auth; tool arguments use country and language.

New

• engine parameter on /v1/search, /v1/news, /v1/images and /v1/videos: google (the default, unchanged), auto, one provider or a list. Opt-in fallback to other search engines when Google is unavailable, labelled in meta.engine and the new meta.route, plus a per-key dashboard setting. See Search providers.
• POST /v1/rank: the best position of a domain for a keyword in the top 10 to 100.
• GET /v1/status: public per-endpoint success rate and latency over the last hour.
• Official SDKs: TypeScript (npm install serpkite), Python (pip install serpkite, sync and async) and Go (github.com/serpkite/serpkite-go), plus langchain-serpkite and serpkite[crewai]. See SDKs.

Unchanged

• format (json, compact, markdown), fields, include_content, num: 100 depth bundle, max_age caching, the AI Overview at no extra cost, resolved destination URLs and the credit prices.
• Response headers X-Credits-Used, X-Credits-Remaining, X-Cost-USD, X-Cache, X-Latency-Ms, X-Tokens-Estimate, X-Request-Id.
• Failed, empty and blocked requests are refunded automatically.
• Account controls: multiple keys with monthly limits, spend cap, alerts, auto-recharge, teams and a metadata-only request log.

v0.9 (2026-09): preview

The first public preview: JSON search endpoints for Google’s verticals, a webpage-to-Markdown fetcher, the CSE-compatible GET /customsearch/v1, the three output formats, the batch lane with signed webhooks, the remote MCP server, and the dashboard with keys, usage, billing and spend controls.

Planned

These are on the roadmap but not live. Nothing here is a date commitment.

• OAuth for the MCP server, so clients that can’t send custom headers (ChatGPT connectors, some desktop apps) can sign in directly.

Incidents

This page covers the API contract only. For outages, degraded performance and incident history, see the status page.