Serper
Serper provides real-time Google search results, place data, autocomplete suggestions, and webpage extraction through a structured API.
Verdict
Common use cases
- Audit competitor landing pages before launch
- Verify customer bug reports with live searches
- Pull news coverage for weekly PR digests
- Find local business reviews for market research
- Scrape product pages for pricing comparisons
Integration
- Vendor
- Serper
- Category
- other
- Auth
- API_KEY
- Tools
- 11
- Composio slug
serper
Tools
- Batch Search the Web
Run 1-10 independent first-page Google web searches in one native Serper request. Each query consumes one credit, and result envelopes preserve input order. Use SERPER_SEARCH_WEB separately when page continuation is needed.
- Get Search Suggestions
Return Google autocomplete suggestions for a partial search query without running a full search.
- List Locations
Find canonical Serper location names and country codes for localizing search requests; this lookup consumes no search credits.
- List Place Reviews
Return paginated Google reviews for one place identified by Place ID, feature ID, or CID, with optional sorting and topic filtering.
- Scrape Webpage
Ask Serper to fetch one webpage and return extracted text, metadata, and optionally markdown, links, images, and videos. Difficult pages may cost 6 or 10 credits instead of 2.
- Search Images
Search Google Images and return one page of source pages, full image URLs, thumbnails, dimensions, and positions. Requesting 100 results costs 2 credits instead of 1.
- Search Maps
Search Google Maps by text, Place ID, or CID, optionally constrained to a viewport, and return one page of rich place identifiers and contact details. Each page costs 3 Serper credits.
- Search News
Search recent Google News coverage and return one page of article titles, links, sources, publication times, snippets, and images when available.
- Search Scholar
Search Google Scholar and return one page of academic publications with citation counts, publication metadata, snippets, and PDF links when available.
- Search Shopping
Search Google Shopping and return one page of products, prices, merchants, offers, ratings, images, and product IDs. Each page costs 2 credits.
- Search the Web
Search Google web results and return one page of organic links plus available answer, knowledge, question, related-search, and top-story sections.
Setup
Setup guide
- 11. Open your Switchy workspace settings and navigate to the Integrations tab. 2. Find Serper in the MCP directory and click Connect. 3. Log into your Serper account at serper.dev, navigate to API Keys, and generate a new key. 4. Paste the key into Switchy's connection modal and click Authorize. 5. Open any Project, type '@Serper search the web for [your query]' in a message, and send. 6. Serper will return first-page Google results inline; if you see structured snippets and links, the connection works. 7. Check your Serper dashboard to confirm credit usage after each search. 8. For batch queries, use '@Serper batch search' and list up to ten queries in one request to save round trips.
What teammates see: by default, memories from Serper are scoped to the Project (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Projects have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Competitor Content Audit
@Serper scrape the homepages for acme.com, widgetco.io, and example.net, then extract their headlines, CTAs, and any feature lists you find.
News Digest for Standup
@Serper search news for 'SaaS collaboration tools' published in the last 24 hours, then list the five most relevant headlines with links.
Local Business Research
@Serper search maps for 'coffee shops in Brooklyn' and return the top five results, then list reviews for each place and summarize recurring themes.
Image Asset Hunt
@Serper search images for 'remote team video call diverse' and return the first 20 results with full image URLs and dimensions.
Autocomplete Keyword Ideas
@Serper get search suggestions for 'how to improve team' and return the full list of autocomplete phrases Google offers.
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@serper search the web for 'MCP server best practices' and 'Claude desktop integration guide' then summarize the top recommendations from both queries
Ran two parallel web searches via Serper's batch endpoint. First query returned 10 results including Anthropic's official MCP documentation, a GitHub discussion on server architecture, and three blog posts covering error handling and transport layers. Second query surfaced the Claude Desktop quickstart, two community tutorials on config.json setup, and a Reddit thread troubleshooting stdio connections. Common themes across both result sets: always validate tool schemas before deployment, use structured logging for debugging, and test with multiple client implementations. Anthropic's docs emphasize security boundaries; community posts highlight the importance of clear tool descriptions for prompt engineering.
Batch search runs up to 10 queries in one request, consuming one credit per query. This example demonstrates pairing Serper's parallel retrieval with the AI's synthesis capability—useful when you need to cross-reference multiple topics quickly. Results reflect first-page Google rankings at query time; no historical or deep-page data.
@serper find the top 5 Google Maps results for 'coffee shops in Seattle' near Pike Place Market, then pull reviews for the highest-rated location
Maps search returned five coffee shops within 0.3 miles of Pike Place Market: Storyville Coffee (4.4★, 890 reviews), Monorail Espresso (4.6★, 340 reviews), Cherry Street Coffee (4.3★, 520 reviews), Uptown Espresso (4.5★, 280 reviews), and Seattle Coffee Works (4.7★, 410 reviews). Fetched 20 most recent reviews for Seattle Coffee Works (Place ID ChIJ...). Reviewers praised the pour-over bar and knowledgeable baristas; three mentioned limited seating. One review flagged inconsistent weekend hours. Average review length: 85 words. Most recent review posted 4 days ago.
Maps search costs 3 credits per page; review retrieval is separate and paginated. This workflow showcases chaining two Serper tools—geographic search followed by reputation deep-dive. Useful for location research, competitor analysis, or travel planning. Review freshness depends on Google's index; sorting options include 'most_relevant', 'newest', 'highest_rating', 'lowest_rating'.
@serper scrape the webpage at anthropic.com/news/model-context-protocol and extract all markdown-formatted links to GitHub repositories or documentation sites
Scraped Anthropic's MCP announcement page (2 credits). Extracted 1,840 words of body text and 14 links. Markdown-formatted repository links: github.com/modelcontextprotocol/servers (official server collection), github.com/modelcontextprotocol/typescript-sdk (TypeScript SDK), github.com/modelcontextprotocol/python-sdk (Python SDK). Documentation links: modelcontextprotocol.io/introduction (spec overview), modelcontextprotocol.io/docs/concepts/transports (transport layer guide). Also found three links to Anthropic blog posts and two to Claude API docs. Page metadata: published October 2024, last modified November 2024, author field empty.
Webpage scraping costs 2 credits for simple pages, up to 10 for JavaScript-heavy or protected sites. Serper returns plain text, markdown, and structured link/image lists—ideal for extracting references from articles or documentation. This example pairs scraping with AI filtering to isolate specific link types. Be aware: some sites block Serper's user agent or require authentication Serper cannot provide.
Use-case deep-dives
When batch search beats manual research for product teams
A 6-person product team preparing quarterly roadmap reviews needs to check competitor feature announcements, pricing changes, and press coverage across eight rivals. The Batch Search the Web tool runs all eight queries in one call, preserving order so each competitor's results stay grouped. Search News pulls recent coverage without opening ten browser tabs. This workflow costs 16 credits total and takes two minutes instead of an hour of tab-switching. The trade-off: if you need results beyond page one or want to track changes daily, you'll burn credits fast—Serper charges per query, so continuous monitoring gets expensive above 200 searches a month. For quarterly or monthly check-ins where you need breadth over depth, this MCP is the right call.
Why Serper works for one-off questions but not help-desk scale
A three-person support team handling 40 tickets a week occasionally needs to verify whether a customer's issue matches a known bug discussed in forums or GitHub. The Search the Web tool finds the thread in seconds, and Scrape Webpage pulls the full discussion without leaving the chat. This costs 3 credits per lookup, manageable at low volume. The problem: if ticket volume climbs to 200 a week and half need external lookups, you're spending 300 credits monthly just on search—Serper's free tier covers 2,500, but you'll hit limits fast. Use this MCP when lookups are exceptions, not the primary workflow. For high-volume support, a dedicated knowledge base or forum integration will cost less and index faster.
When Maps search beats LinkedIn for territory planning
A two-person sales team targeting dental practices in metro Phoenix needs contact details and review sentiment before cold outreach. Search Maps returns 20 practices per query with phone numbers, addresses, and ratings; List Place Reviews pulls patient feedback to gauge reputation. Each Maps search costs 3 credits, so mapping 100 prospects costs 15 credits total—cheaper than buying a lead list and faster than scraping LinkedIn. The boundary: Serper returns what Google Maps shows publicly, so if your vertical hides contact info or lives outside Google's index (niche B2B services, private practices), you'll get incomplete data. For consumer-facing local businesses with strong Google presence, this MCP closes the research loop in minutes.
Frequently asked
What does the Serper MCP let me do in Switchy?
It runs Google searches—web, images, news, and maps—directly inside your Switchy workspace without opening a browser. You can batch up to ten queries at once, scrape webpages for clean text, pull autocomplete suggestions, and fetch Google Maps reviews. Every result stays in your conversation history so the whole team can reference it.
Do I need a Serper account to connect this MCP?
Yes. You'll need a Serper API key, which means signing up at serper.dev and choosing a credit plan. Switchy passes your key with every request; there's no OAuth dance. Keep the key private—anyone with it can burn through your Serper credits.
Can the Serper MCP search beyond the first page of results?
No. Every tool returns one page of results—typically ten items for web search, up to one hundred for images if you pay double credits. If you need deep pagination, you'll have to call the Serper API directly or run multiple searches with different date filters.
Why use this instead of just Googling things myself?
Speed and context. The MCP pulls structured JSON instead of HTML, so you can feed results straight into another AI task—summarize ten news articles, compare product pages, extract contact details from Maps. You skip the copy-paste loop, and every search becomes a reusable artifact in your workspace.
Does Serper usage count against my Switchy plan limits?
No. Serper charges you separately by credit—one credit per web search, two for a scraped page, three for Maps. Switchy doesn't meter those calls. Just watch your Serper dashboard so you don't blow through your monthly allowance mid-project.