# AddSearch MCP for AI Agents MCP > AddSearch gives your AI agent direct access to your site's search index. You can run deep queries using natural language, manage documents by listing or adding content via JSON, and retrieve live search performance analytics immediately. It lets you audit site search relevance without needing complex dashboards. ## Overview - **Category:** developer-tools - **Price:** Free - **Tags:** site-search, nlp-search, search-analytics, api-indexing, content-discovery, search-ranking ## Description This MCP connects your search index directly into your AI workflow, turning what used to be a complicated backend database into something usable through chat. Instead of jumping between multiple tabs or running specific scripts, you just ask your agent questions about your site's performance or content structure. For instance, you can ask which product categories are generating the most zero-result queries—a massive time saver for any content team. It also lets developers debug the index by manually adding a fixed document or checking metadata on a single URL. This kind of deep, structured access to search data is what makes Vinkius such a valuable hub; you connect once and instantly gain these powerful capabilities across all your AI tools. ## Tools ### delete_document Permanently removes a specified document from your site index. Requires the secret key. ### search_filtered Searches indexed content by applying custom field filters, such as looking only for 'category=shoes' or 'brand=nike'. ### index_document Adds a brand-new document to your site index or updates an existing one. Requires the secret key. ### list_documents Lists all documents currently indexed on your site. Requires the secret key. ### search_pagination Retrieves a specific page number of search results for deep content review. ### search_keyword Performs a basic keyword search across your entire indexed site content. ### search_sorted Searches and ranks indexed content using custom sorting criteria, allowing for targeted result sets. ### stats_clicks Retrieves detailed analytics on how many times search results were clicked by users. Requires the secret key. ### stats_queries Gathers comprehensive data on all user searches performed, helping identify popular and forgotten topics. Requires the secret key. ### autosuggest Generates a list of autocomplete suggestions based on common prefixes typed into your search bar. ## Prompt Examples **Prompt:** ``` Show me all product pages for 'hiking boots' that are currently marked as discontinued. ``` **Response:** ``` **🔍 Search Results: Discontinued Items** | Product ID | Name | Last Indexed Date | Status | | :--- | :--- | :--- | :--- | | P1045 | Trail Blazer Elite | 2023-01-15 | Archived | | P987 | Summit Grip GTX | 2023-04-01 | Discontinued | *Recommendation: These pages should be redirected to our 'Best Sellers' category.* ``` **Prompt:** ``` What are the most common search terms that users type in but find nothing? ``` **Response:** ``` **⚠️ Top Zero-Result Queries (Last 7 Days)** The data shows a clear pattern of missed opportunities: * 'corporate retreat planning': 45 searches * 'advanced Python class': 28 searches * 'gluten free wedding cakes': 19 searches Focusing new content on these topics will likely boost engagement. ``` **Prompt:** ``` Test the auto-suggest for 'marketing strategy'. ``` **Response:** ``` **✨ Auto-Suggest Suggestions Found:** The search bar is ready to guide users with these completions: * marketing strategy checklist * B2B marketing automation * digital marketing trends 2026 This confirms the index successfully captured all related terms for immediate suggestion. ``` ## Capabilities ### Search Indexed Content by Query You can query your entire site content using plain English, or narrow the search down using specific field filters like a category name. ### Analyze Search Failure Points Retrieve statistics that show which user searches failed to find results, helping you prioritize new content creation. ### Manage Indexed Documents List every document in your index or permanently delete outdated pages using simple commands. ### Push Content Updates Add new articles or update existing documents directly to the search index using a JSON structure. ### Test Search User Experience Simulate how a real user interacts with your site's search bar, testing auto-suggestions and pagination links. ## Use Cases ### Identifying Content Gaps from Failed Searches A Content Manager needs to know why users are searching for 'remote coding jobs' but getting zero results. They ask their agent, which uses stats_queries, and immediately get a list of the top 10 failed queries, giving them an actionable content backlog. ### Debugging Product Relevance in E-commerce An E-Commerce Specialist runs a test query: 'leather boots' filtered by category=seasonal. The agent uses search_filtered to verify if the ranking correctly shows only the current inventory, confirming relevance before launch. ### Cleaning Up Outdated Site Content A Developer needs to remove a deprecated product line from the index immediately. They instruct their agent to use delete_document with the specific URL ID, ensuring the content is gone instantly across all search results. ### Validating Search Bar Functionality A UX designer tests the site's front end by asking the agent to run autosuggest for 'sustainable fabric'. The agent returns the top suggestions, letting the designer confirm if the index is capturing niche terms correctly. ## Benefits - Stop guessing about site performance. By using stats_queries, you instantly find out what searches are failing to return results, letting your agent guide content creation where it matters most. - Debugging is faster. Instead of going into the backend, you can use index_document or delete_document to fix metadata or remove outdated pages directly through chat when necessary. - Deep filtering lets you drill down past basic searches. The search_filtered tool means you can ask for 'all shoes in the men's section from brand X,' getting surgical precision every time. - Understand user behavior with clicks. By retrieving click-through analytics via stats_clicks, your agent shows you which results are actually valuable to users, not just what exists in the index. - Testing is built in. Use the search_pagination tool to test how deep your site's content goes and verify that auto-suggestions work properly for any given prefix. ## How It Works The bottom line is you tell your AI agent what data point about the index you need, and it fetches it instantly without you touching any dashboards. 1. Subscribe to the AddSearch MCP and enter your Site Key into your agent environment. 2. If needed, input your optional Secret Key for advanced operations like deleting or adding documents. 3. Your AI client uses these tools to execute complex search queries and retrieve structured site analytics on demand. ## Frequently Asked Questions **How do I find out what content pages on my site are getting zero search results?** You can use the AddSearch MCP to retrieve detailed analytics showing exactly which user queries failed. This is perfect for identifying critical content gaps that need to be written or improved immediately. **Can I update product information in my site's search index using this tool?** Yes, you can use the AddSearch MCP's document management tools. By sending a new JSON payload, your agent updates an existing page or adds a brand-new one instantly to the live search index. **Is AddSearch better than just looking at Google Analytics?** Yes. While GA shows traffic flow, this MCP gives you direct access to the internal metadata and structured results of your site's own search engine, letting you debug relevance issues others can't see. **What if I need to check a specific category or product line?** You use the AddSearch MCP’s filtered search tools. You tell your agent exactly what criteria to apply—like 'category=electronics and brand=apple'—to get highly targeted results. **How do I check if my auto-suggestions are working correctly?** You run a test using the AddSearch MCP. It simulates typing into your search bar, returning all possible autocomplete suggestions based on the current index data so you can verify functionality.