# Marqo AI MCP for AI Agents AI Agent Connect

> Marqo AI (Vector Search & Embeddings) MCP lets you manage your semantic search infrastructure through your AI agent. You can execute tensor queries, index JSON documents, and audit vector indices in real-time. It connects your Marqo instance to any MCP-compatible client so you can handle embeddings and document lifecycles without leaving your workspace.

## Overview
- **Category:** ai-frontier
- **Price:** Free
- **Endpoint:** https://edge.vinkius.com/vk_preview_yWwQAD75WfnpbXY72FMh8SBRNTeZWrryXTJG2Ifa/ai-agent-connect
- **Tags:** semantic-search, vector-embeddings, tensor-search, indexing, information-retrieval

## Description

Imagine you're trying to build a search feature that actually understands what a user means rather than just matching keywords. You're stuck jumping between a web console, your IDE, and your database to check if your embeddings are landing right or if your index is getting bloated. This Connector changes that by bringing your Marqo instance directly into your AI agent's reach. Instead of manual API calls, you can just tell your agent to find specific items using natural language, and it handles the tensor search logic for you. You can also feed new data into your vector indices on the fly, making sure your search results stay fresh without a manual re-indexing headache. If you need to clean up old data or check the health of a specific collection, you just ask. It's a way to manage your entire vector search architecture as part of your normal conversation with your agent. It makes sense to have this kind of control when you're scaling complex data, and it's one of the reasons why having a managed catalog like Vinkius is so helpful for keeping these connections organized and ready to go.

## Tools

### get_index_stats
Pull the configuration and stats for a specific index. Use it to check your document counts and embedding model types.

### tensor_search
Execute a natural language tensor search on your data. The agent handles the embedding extraction so you just get the results.

### add_documents
Write new documents into your Marqo vector indices. This lets you update your searchable data in real-time.

### delete_documents
Remove specific documents from Marqo using their unique IDs. It keeps your search index clean and relevant.

### create_index
Create a new vector index with specific bounds and settings. This helps you set up new search architectures on the fly.

### list_indexes
See all your Marqo vector indexes at once. This helps you identify which collections are available before you run any queries.

## Prompt Examples

**Prompt:** 
```
Check the status of my 'support-docs' index.
```

**Response:** 
```
I've pulled the stats for your 'support-docs' index. Here's what I found:

| Metric | Value |
| :--- | :--- |
| Document Count | 1,240 |
| Embedding Model | marqo-v3-small |
| Dimensions | 768 |

Everything looks healthy. Would you like me to run a test search on it?
```

**Prompt:** 
```
Add this to the 'products' index: {'id': 'p101', 'name': 'Trail Runner', 'desc': 'Lightweight shoes'}
```

**Response:** 
```
I've added that product to your 'products' index. Marqo has vectorized the description and the record is now searchable. Your new document ID is `marqo-99283`.
```

**Prompt:** 
```
Find some trail running shoes in the products index.
```

**Response:** 
```
I'm running a tensor search for 'trail running shoes' in your 'products' index now...

I found 3 highly relevant matches:
* **Swift-Trail Runner** (Score: 0.94)
* **Mountain Glide** (Score: 0.88)
* **Peak Hiker** (Score: 0.82)

Would you like to see the full metadata for any of these?
```

## Capabilities

### Run natural language tensor searches
Ask your agent to find specific items and it handles the embedding extraction automatically.

### Add new JSON documents
Push new data into your vector indices to keep your search results up to date.

### Create new vector indices
Set up new search collections with specific bounds and model settings on the fly.

### View index statistics
Check your document counts and embedding model types to monitor your data health.

### List all available indices
See every vector index in your Marqo instance before you start running queries.

### Delete specific documents
Remove old or incorrect records by targeting their unique IDs to keep your index clean.

## Use Cases

### Updating a product catalog
A developer needs to add 50 new items to a search index. They ask the agent to push the JSON data into Marqo instantly.

### Debugging search relevance
A search architect isn't happy with results. They ask the agent to run a tensor search on a specific query to see what the top matches are.

### Cleaning up stale data
An engineer needs to remove old user profiles. They tell the agent to clear out specific IDs from the vector index.

### Setting up a new search space
A team needs a new index for technical docs. They ask the agent to set up a bounded index with specific dimensions.

## Benefits

- Stop switching tabs by managing your Marqo vector indices directly within your AI client.
- Keep your search results fresh by pushing new data into your vector indices in real-time.
- Verify your search accuracy faster by running tensor searches using natural language.
- Audit your data health easily by checking your document counts and model types.
- Organize your search architecture better by setting up new indices with custom bounds.
- Maintain a clean index by removing specific records by their unique IDs.

## How It Works

The bottom line is you get a direct line of communication between your AI agent and your Marqo vector database.

1. Subscribe to the Marqo AI MCP on Vinkius.
2. Enter your Marqo API URL and API Key in your client settings.
3. Start asking your agent to search, index, or manage your vector data.

## Frequently Asked Questions

**How does the Marqo AI MCP help with my search results?**
It lets your AI agent perform tensor searches directly on your Marqo data. Instead of just matching words, it understands the meaning behind your queries to give you better results.

**Can I use Marqo AI to add new data to my vector database?**
Yes. You can tell your agent to add new JSON documents to your Marqo indices whenever you need to. It handles the indexing for you automatically.

**Is the Marqo AI MCP good for managing large vector indices?**
It's built for that. You can use it to list all your indices, check their stats, and monitor your document counts to keep your search infrastructure organized.

**How do I delete old records from my Marqo instance?**
You can ask your agent to remove specific documents by their unique IDs. This helps keep your search index clean and relevant.

**Can I create new search collections with this Connector?**
Yes, you can use it to create new vector indices with specific dimensions and model settings. It's a fast way to expand your search capabilities.

**Does the Marqo AI MCP work with my existing Marqo account?**
It does. You just need to provide your Marqo API URL and API Key to connect your existing instance to your AI agent.

**Does Marqo handle the vector embeddings for me through the agent?**
Yes. Marqo is an end-to-end engine. When you use the `tensor_search` tool, you provide natural language and Marqo handles the model inference and vector extraction under the hood, returning semantically relevant results immediately.

**Can I add new data to a vector index through a conversation?**
Absolutely. Use the `add_documents` tool by providing a JSON array of your documents. Your agent will synchronize these records into the target index, and they will be searchable via semantic query instantly.

**How do I check the stats of my vector index?**
The `get_index_stats` tool retrieves critical metrics for a specific index. Your agent will report the document count, memory usage, and configuration details, helping you monitor the health of your vector store.