# SerpApi

Scrapes Google, Bing and other search engines on request and returns the result pages as structured JSON or Markdown.

- Provider: SerpApi, LLC (https://serpapi.com)
- Category: Search the web
- Type: Hosted API
- Pricing model: Monthly subscription by search count; cancel anytime [1]
- Free tier: 250 searches per month [1]
- Authentication: API key (`api_key` parameter; `Authorization: Bearer` header for the MCP server) [3]
- Connect via: REST API, MCP server, SDKs, CLI [5]
- Read from the provider's pages on 2026-11-04
- Page: https://supermarket.dev/tools/search-the-web/serpapi

## Pricing

### Billing
- Pricing model: Monthly subscription with a fixed number of searches; unused plan can be topped up by early renewal [1]
- Free tier: 250 searches per month, 50 searches per hour guaranteed [1]
- Card required for free tier: Not published
- Minimum spend: $0 on the Free plan; $25 per month on the lowest paid plan [1]
- Billing period: Monthly [1]

### Prices
Per 1,000 searches = monthly price divided by searches included (Starter 25/1; Developer 75/5; Production 150/15; Big Data 275/30). Rounded to cents.
| Plan | Monthly price | Searches per month | $ per 1,000 searches |
|---|---|---|---|
| Free | $0 | 250 | $0 |
| Starter | $25 | 1,000 | $25.00 |
| Developer | $75 | 5,000 | $15.00 |
| Production | $150 | 15,000 | $10.00 |
| Big Data | $275 | 30,000 | $9.17 |
| Enterprise | Contact sales | Custom | Custom |
Source [1]

### Other charges
- What counts as a search: Only successful, uncached searches; result count does not change the count [1]
- Cached searches: Free, not counted [4]
- Running out of searches: Automatic Early Renewal or manual renewal [1]
- Refund: Full refund within 7 days unless more than 20% of searches used [1]
- Crawling with Extract: 1 search credit per extracted page on an uncached request [3]

## Tools

### MCP server
- Official MCP server: Yes, remote and local [3]
- Address or package: `https://mcp.serpapi.com/mcp` (remote); `https://github.com/serpapi/serpapi-mcp` (self-host, stdio `.mcpb` bundle) [3]

### MCP tools
From tools/list on the remote server, no credentials, 4 tools.
| Tool | What it does | Read-only |
|---|---|---|
| `search` | Universal search over all SerpApi engines and result types | Yes |
| `extract` | Extracts content from a public URL, maps site URLs, or crawls pages | Yes |
| `search_table` | Same params as `search`; shows organic results as an interactive table | Yes |
| `search_dashboard` | Same params as `search`; shows metrics, source chart and results table | Yes |
Source [11]

### API
| Endpoint | What it does |
|---|---|
| `GET https://serpapi.com/search` with `engine=google` | Google web results as JSON; `engine` selects other engines such as `bing`, `duckduckgo`, `google_news` |
| `GET https://serpapi.com/search` with `engine=google_light` | Faster Google Search variant |
| `GET https://serpapi.com/search` with `engine=google_ai_mode` | Google AI Mode results |
| `GET https://serpapi.com/search` with `engine=extract` | Reads a page as structured JSON or Markdown; scrape, map or crawl modes |
| `/locations.json` | Looks up location names for the `location` parameter |
| Searches Archive API | Retrieves stored or async searches later |
Source [4]

### Features
- Results per query: Not published
- Content returned: Structured JSON (default), raw HTML, or Markdown for LLMs [4]
- Search types: Web, images, local, videos, news, shopping, patents, books, AI Mode, short videos, forums [4]
- Freshness controls: `tbs` advanced filters (dates among them); `no_cache` forces a fresh fetch, cache expires after 1h [4]
- Domain filters: `site:` and other operators in `q` [4]
- Storing results: Not published

## Authentication

### Methods
| Method | Used for | How |
|---|---|---|
| API key in query | REST API | `api_key` parameter on every request |
| Bearer header | MCP server | `Authorization: Bearer YOUR_API_KEY` |
| Key in URL path | MCP clients that cannot set headers | `https://mcp.serpapi.com/YOUR_API_KEY/mcp` |
| None | Connecting to the MCP server, listing tools, reading resources | `search` and the app tools need a key |
Source [3]

### Getting access
- Who creates the account: A person signs up; no agent self-signup stated [4]
- Where to get a key: Dashboard key page `https://serpapi.com/manage-api-key` [3]
- Other requirements: Not published

## Setup

### Addresses
- REST API: `https://serpapi.com/search` [4]
- MCP server: `https://mcp.serpapi.com/mcp` [3]
- OpenAPI spec: Not published
- Docs for agents: `https://serpapi.com/llms.txt`; docs pages also served as `.md` [6]
- Status page: `https://status.serpapi.com` [10]

### Install
| What | Command |
|---|---|
| Claude Code (MCP) | `claude mcp add --transport http serpapi https://mcp.serpapi.com/mcp --header "Authorization: Bearer YOUR_SERPAPI_API_KEY"` |
| Codex (MCP) | `codex mcp add serpapi --url https://mcp.serpapi.com/mcp --bearer-token-env-var SERPAPI_API_KEY` |
| Python SDK | `pip install serpapi` |
| JavaScript SDK | `npm install serpapi` |
| CLI (Homebrew) | `brew tap serpapi/homebrew-tap && brew install serpapi-cli` |
| MCP self-host | `git clone https://github.com/serpapi/serpapi-mcp.git && cd serpapi-mcp && uv sync && uv run src/server.py` |
Source [3]

## Limits

### Rate limits
| Plan | Guaranteed successful searches per hour |
|---|---|
| Free | 50 |
| Starter | 200 |
| Developer | 1,000 |
| Production | 3,000 |
| Big Data | 6,000 |
| Enterprise | Custom |
Source [1]

### Other limits
- Monthly quota: Set by plan: 250 to 30,000 searches; custom on Enterprise [1]
- Crawl size: Default 10 pages, maximum 100 per crawl [3]
- Cache window: Identical searches served from cache for 1 hour [4]
- Uptime SLA: Up to 99.97% on premium support [1]
- Concurrency: Not published

## Privacy

### Your data
- Data location: United States [7]
- Retention: Search data deleted automatically 31 days after the search [7]
- Used for training: Not published
- Zero data retention: Yes on Enterprise plans: ZeroTrace Mode with `zero_trace=true` [9]

### Compliance
- Certifications: SOC 2 Type II, SOC 3, ISO 27001 (ISO 27701 and EWDCI seal also listed as acquired) [1]
- HIPAA: No HIPAA attestation stated; ZeroTrace page says it can make HIPAA compliance easier [9]

## Sources

1. Plans and pricing: https://serpapi.com/pricing
2. FAQ: https://serpapi.com/faq
3. MCP integration: https://serpapi.com/integrations/mcp
4. Google Search API (Markdown): https://serpapi.com/search-api.md
5. Integrations (Markdown): https://serpapi.com/integrations.md
6. llms.txt: https://serpapi.com/llms.txt
7. Legal: terms and privacy: https://serpapi.com/legal
8. Security: https://serpapi.com/security
9. ZeroTrace Mode: https://serpapi.com/zero-trace-mode
10. API status: https://serpapi.com/status
11. Checked by supermarket.dev