# Vald MCP for AI Agents AI Agent Connect > Vald MCP gives your agent direct access to a distributed vector search engine. It lets you query, insert, and manage high-dimensional embeddings for RAG pipelines and similarity searches. Instead of writing complex scripts to talk to your vector cluster, your agent handles the heavy lifting of finding the nearest neighbors in your data. ## Overview - **Category:** databases - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_LwabvjFRxR6SLDMyq0Lg8wB1fanBWcNw0aRuqiDM/ai-agent-connect - **Tags:** vector-search, ann, distributed-systems, machine-learning, high-performance, embeddings ## Description You're trying to build a RAG pipeline, but getting your agent to actually find the right context out of millions of data points is a pain. This Connector changes that by putting your Vald cluster front and center. You can stop jumping between your terminal and your IDE to check if a vector is actually in the index or to see if your cluster is healthy. From here, your agent can pull raw vector data, shove new embeddings into the system, or wipe out corrupted records without you having to touch a single line of boilerplate code. It turns your conversational interface into a control plane for your vector database. If you're looking for a way to manage distributed nearest-neighbor search without the friction of manual database queries, this is how you do it. You can find this listed in the Vinkius catalog to get your agent connected in minutes. ## Tools ### insert_vector Add a new vector and its unique ID into your Vald index. This is the primary way to ingest new data points. ### delete_vector Permanently remove a specific vector from the cluster. It is useful for cleaning up your search space. ### get_engine_info Get the current health and operational status of your Vald engine. This helps you monitor cluster stability. ### get_vector_details Pull the raw float array for a specific vector ID to check its contents. Use this to verify your embedding logic. ### update_vector Replace an existing vector's data with a new array using its ID. This keeps your records current without re-indexing. ### search_vectors Run a nearest neighbor similarity search using a query vector. This is what powers your semantic search results. ## Prompt Examples **Prompt:** ``` Is the Vald cluster healthy right now? ``` **Response:** ``` I've checked the Vald Gateway. The engine is healthy and accepting requests! The current clustered infrastructure responds well to Ping mechanisms. ``` **Prompt:** ``` Can you check the vector details stored for 'user-profile-89'? ``` **Response:** ``` I retrieved the vector data for `user-profile-89`. The embedding array spans 768 dimensions. Do you want me to render its raw float values here or proceed to query for its nearest neighbors? ``` **Prompt:** ``` Update the existing item 'context-fragment-12' with this new 1536-dimensional array: [0.38, -0.19, 0...]. ``` **Response:** ``` Done. The vector for ID `context-fragment-12` has been successfully updated in your Vald cluster. The engine replaced its old representation with your newly uploaded 1536-dimensional array without breaking any active query connections. ``` ## Capabilities ### Perform similarity searches Query your embeddings to find the most relevant data points instantly. ### Add new data points Insert high-dimensional vectors into your active index for immediate use. ### Refresh existing records Update specific vector representations without rebuilding the entire index. ### Remove old data Permanently delete vectors from the cluster to keep your search space clean. ### Monitor cluster health Get a real-time look at your engine's status and node details. ### Inspect raw data Pull specific vector values by ID to verify your embedding pipeline. ## Use Cases ### Debugging a broken RAG pipeline A developer notices the AI is pulling irrelevant facts. They ask the agent to check the raw data for a specific ID using get_vector_details to see if the embedding is actually correct. ### Rapidly testing new embeddings An ML engineer wants to see how new model outputs look in the cluster. They ask the agent to use insert_vector to add a batch of new points and then run a search. ### Handling production errors A DevOps engineer sees a spike in latency. They ask the agent to run get_engine_info to see if any nodes in the Vald cluster are down or struggling. ### Cleaning up legacy data A backend dev needs to remove old user profiles. They tell the agent to use delete_vector to purge the old entries from the nearest-neighbor engine. ## Benefits - Skip the boilerplate: Use insert_vector and update_vector to manage your data without writing custom API calls every time. - Faster debugging: Use get_vector_details to see exactly what's in your index during the development phase. - Proactive monitoring: Keep an eye on your cluster health with get_engine_info to catch issues before they hit production. - Real-time search: Perform similarity searches with search_vectors to test your RAG recall instantly. - Cleaner indices: Use delete_vector to remove old or broken data points without manual database intervention. ## How It Works The bottom line is you get a direct line between your AI client and your distributed vector search engine. 1. Subscribe to the Vald MCP on Vinkius. 2. Input your Vald Gateway Host address into your AI client. 3. Ask your agent to search, insert, or check the health of your vector cluster. ## Frequently Asked Questions **How does the Vald MCP help with RAG pipelines?** It gives your AI agent a direct way to interact with your vector database. You can search for relevant context, insert new data points, and manage your embeddings without writing any extra code. **Can I use Vald MCP to check if my vector cluster is online?** Yes. You can ask your agent to check the health of your Vald cluster at any time. It will pull the latest status from your Gateway so you know the engine is running smoothly. **Can I update specific vectors using the Vald MCP?** You can. If you have a specific ID, you can tell your agent to update that vector with a new array of floats. It handles the update process across your distributed cluster automatically. **Is it possible to delete vectors with the Vald MCP?** Yes, you can permanently remove specific vectors from your index. This is useful for cleaning up corrupted data or removing old records to keep your search results accurate. **How does the Vald MCP handle high-dimensional embeddings?** It's designed specifically for that. It lets your agent perform nearest neighbor searches across millions of high-dimensional points, making it ideal for complex similarity tasks. **Can I see the raw data of a vector using Vald MCP?** You can request the raw vector details for any specific ID. Your agent will pull the float array from the Vald index so you can verify the exact values stored in your database. **Can my AI agent do a semantic search across my vector database?** Yes! Provided you supply the embedded query vector, your agent can issue a `vector search` command to the Vald Engine. It will rapidly scan millions of indexes natively using its ANN algorithms and return the top-K closest neighbors associated with your data. **How do I ensure my Vald cluster is healthy right from my CLI?** Skip complex diagnostics loops. Instruct your agent to `get Vald internal engine info`. It will interface directly via gRPC/REST and pull down cluster metrics including operational status, agent versions, and basic diagnostic health. This is vital for MLOps managing production RAG pipelines needing constant reassurance. **Can I permanently purge a corrupted vector embedding?** When a document becomes stale in your knowledge base, you must remove its embedding. Ask the AI agent: `permanently delete vector ID 'doc-xyz'`. Using the `removeVector` capability, it targets your cluster and ensures the outdated semantic representation is fully expunged without risking other node data.