# Lucidworks Fusion MCP for AI Agents AI Agent Connect > Lucidworks Fusion MCP gives your AI agent direct control over your enterprise search architecture. Use it to run semantic queries, manage document indexing, and monitor ML training jobs in one place. It bridges the gap between your agent and your Lucidworks Fusion instance, letting you audit pipelines and inject user signals to improve search relevance without leaving your chat interface. ## Overview - **Category:** knowledge-management - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_jD2MOpvp0XAN4LgOxrkdAYw47AZunMSJvOVlZSCX/ai-agent-connect - **Tags:** enterprise-search, semantic-search, machine-learning, vector-search, data-ingestion, search-ranking ## Description Lucidworks Fusion connects your AI agent directly to your enterprise search architecture. Instead of jumping between different tabs or manual consoles, you can manage your search results, ML ranking, and data ingestion through simple conversation. You'll find it's a lot easier to spot a bottleneck in your query routing or verify that your conversion signals are hitting the right models without leaving your chat interface. Because Vinkius hosts this in its catalog, you get a unified way to manage these high-level search operations alongside hundreds of other tools. It takes the friction out of search engineering by turning complex API calls into natural language. You can ask your agent to pull specific search results, check if your ML models are actually training correctly, or update your document index on the fly. It's about moving from how do I run this query to what are the results for this user behavior and getting the answer in seconds. This means you can spend your time on the high-level strategy of improving user discovery rather than the tedious mechanics of manual data entry or checking log files. You can audit your entire pipeline, from the initial signal ingestion to the final ranking output, and see exactly how your rules are performing. It gives you a way to interact with your search platform as if you were talking to the system itself, making it much faster to iterate on your results and keep your data accurate. ## Tools ### lw.list_collections View all active search collections and their rules. This helps you audit your data distribution across your tenant. ### lw.post_custom_query Send custom JSON logic to override default vector mappings. This is useful for specific overrides where standard queries aren't enough. ### lw.query_filtered Extract specific properties that drive your account logic. It helps you pinpoint the exact data points influencing your results. ### lw.list_index_profiles See the active arrays used for parsing your data. This is useful for auditing how your search profiles are structured. ### lw.index_documents Add new documents to your search index immediately. It is the fastest way to keep your data current. ### lw.list_jobs See the status of all active ML and data ingestion jobs. Use this to keep an eye on your backend processing. ### lw.list_query_profiles Run an automated check on your query routing history. It helps you verify that your gateway logic is working. ### lw.query_search Run a standard search to find specific records. It resolves vector rules against your strict profile logic. ### lw.post_signal Send user behavior data like clicks to your search platform. This is how you train your models to be smarter. ### lw.query_sorted Fetch a list of records sorted by date or other criteria. Use this to get a clean, ordered JSON payload of your data. ## Prompt Examples **Prompt:** ``` Can you check if the ML training jobs for our Commerce app are still running? ``` **Response:** ``` I've checked the status for 'Commerce'. You have 2 active jobs: * **Query-Intent-Model**: Currently running (45% complete) * **Click-Re-Ranking**: Currently in a waiting state I'll keep an eye on it for you. Would you like me to check the index profiles as well? ``` **Prompt:** ``` Run a search for 'password reset' in the Support app using the default profile. ``` **Response:** ``` Searching the 'Support' app... I found 5 relevant documents. **Top Result:** 'How to reset your corporate password' (Score: 0.95) Other matches include troubleshooting steps and security policies. Would you like to see the full content of any of these? ``` **Prompt:** ``` I need to add a new document to the index. Here is the text: [Text] ``` **Response:** ``` I've successfully pushed that new content to your search index. It is now available for discovery across your Fusion tenant. Is there anything else you'd like to update? ``` ## Capabilities ### Run semantic and keyword searches Execute complex queries across different apps and profiles to see exactly what your users see. ### Update document indexes Sync new data or update existing records to keep your search results fresh and accurate. ### Monitor ML training Check the status of active training jobs to ensure your ranking models are progressing as expected. ### Inject user signals Send click and conversion data to your search platform to help your ML models learn faster. ### Audit query profiles Inspect the transformation rules and routing logic that determine how your search results are ranked. ### List search collections See all your active collections and physical shards to audit data distribution across your tenant. ## Use Cases ### Testing query profiles A search engineer wants to see how a new query profile handles password reset requests. They ask the agent to run lw.query_search and compare results. ### Monitoring ML job status A data scientist needs to know if the Click-Re-Ranking job is done. They ask the agent to check lw.list_jobs for the Commerce app. ### Updating live content A content manager needs to add a new support doc. They tell the agent to use lw.index_documents to push the new text to the index. ### Auditing search rules A digital experience lead wants to see how query routing is configured. They ask the agent to list the profiles using lw.list_query_profiles. ## Benefits - Faster search testing: Use lw.query_search and lw.query_filtered to test different search behaviors in seconds rather than building new query profiles manually. - Real-time ML monitoring: Keep an eye on your ranking models with lw.list_jobs so you know exactly when your training is finished or if a job fails. - Dynamic data updates: Keep your index fresh by using lw.index_documents to push new content to your search collections instantly. - Improved search relevance: Use lw.post_signal to feed user behavior back into Fusion, allowing your ML models to learn from actual clicks and conversions. - Simplified pipeline auditing: Quickly inspect your routing logic using lw.list_query_profiles to ensure your search rules are behaving as intended. - Clearer data distribution: Use lw.list_collections to get a high-level view of your physical shards and data organization across your entire tenant. ## How It Works The bottom line is you get a conversational interface for managing complex enterprise search infrastructure. 1. Connect your Lucidworks Host URL and API Token to the Connector. 2. Describe the search query or maintenance task to your AI agent. 3. Receive real-time data, job statuses, or confirmation of successful indexing. ## Frequently Asked Questions **What does the Lucidworks Fusion MCP actually do?** It connects your AI agent to your Lucidworks Fusion environment so you can manage search results, monitor ML models, and update your data index using natural language. **Can I use the Lucidworks Fusion MCP to improve my search ranking?** Yes, you can use it to send user behavior signals back to Fusion, which helps your ML models learn from clicks and conversions to improve future results. **How does the Lucidworks Fusion MCP help with data indexing?** It allows your agent to sync new documents or update existing records in your physical collections, keeping your search results fresh without manual entry. **Is the Lucidworks Fusion MCP good for monitoring ML jobs?** It is perfect for that. You can ask your agent to list all active training jobs and see their current status, percentage completion, and any errors in real-time. **Can I audit my search query profiles with Lucidworks Fusion MCP?** Yes, you can use the agent to list and inspect your query profiles to see exactly how your routing rules and transformation logic are configured. **Does the Lucidworks Fusion MCP support semantic search?** It gives your agent the ability to execute complex keyword and vector-based queries against your specific apps and profiles. **Can my agent help improve search relevance by posting signals?** Yes. Use the `lw.post_signal` tool to feed conversion or click events into Fusion. These signals are used by the machine learning models to identify high-value results and re-rank them for better precision automatically. **How do I check which AI models are active in my query pipeline?** The `lw.list_query_profiles` tool retrieves all defined profiles in your gateway. Your agent will expose the exact AI models and routing configurations assigned to each profile, helping you audit your search logic. **Can I index new documents directly through a conversation?** Absolutely. Use the `lw.index_documents` tool by providing a JSON array of your data. Your agent will synchronize these new mappings into your Fusion collection, making the content queryable instantly.