How to Use the NVIDIA Vision MCP in LlamaIndex

Q: How do I handle image URLs in a LlamaIndex query using NVIDIA Vision?

You pass the public image URL directly to the visualquestionanswering tool within your agent's query loop. The tool returns the answer, which LlamaIndex can then use to synthesize a final response to the user.

Index NVIDIA Vision outputs directly into your LlamaIndex vector stores to search and query your visual data.

See Vinkius in Action

Works with every AI agent you already use

…and any MCP-compatible client

MCP Servers - Free for Subscribers

Connect NVIDIA Vision MCP to LlamaIndex

Create your Vinkius account to connect NVIDIA Vision to LlamaIndex and route execution through our secure gateway. The platform manages server hosting, runtime updates, and security layers. Configuration requires no manual server provisioning.

GDPR Free for Subscribers

Setup NVIDIA Vision with LlamaIndex

Ask AI about this MCP

ChatGPT

Claude

Perplexity

Semantic Search Over Image Captions in LlamaIndex

The NVIDIA Vision MCP Server provides the `image_captioning` tool to convert raw visual data into searchable text descriptions that LlamaIndex can index. Your ingestion pipeline sends images to the tool, gets detailed textual descriptions, and writes them directly to your vector database. This eliminates manual tagging. Users can then query your document store using natural language, and LlamaIndex retrieves the relevant images based on the semantic meaning of the generated captions rather than generic file names.

Visual RAG for Scanned Documents

This MCP Server integrates `document_qa` to let your LlamaIndex agents query scanned PDFs, invoices, and receipts directly. Instead of running a separate OCR step, the agent queries the document image and indexes the factual answers alongside your other text documents. This creates a unified knowledge base. Your RAG applications can pull context from both raw text files and scanned images, ensuring your agent has access to all company data during a query session.

Image Segmentation Metadata Indexing

Extracting structured metadata from your image libraries is handled by `image_segmentation` and `detect_objects` via the MCP Server. Your LlamaIndex pipeline uses these tools to identify every object and its location within an image, saving these details as node metadata. When a user searches for specific items, LlamaIndex filters the vector search using this metadata. This ensures highly accurate image retrieval based on actual visual contents rather than automated guesses.

Setup guide

Set up NVIDIA Vision MCP in LlamaIndex

Prerequisites

Python 3.10+ installed
llama-index-tools-mcp package
Active Vinkius subscription with a valid endpoint token

1

Install dependencies
Run pip install llama-index-tools-mcp llama-index-llms-openai. The MCP tools package provides BasicMCPClient and McpToolSpec.
2

Connect with BasicMCPClient
Point BasicMCPClient to your Vinkius endpoint URL. Replace [YOUR_TOKEN_HERE] with your token from cloud.vinkius.com. Supports SSE and Streamable HTTP transports.
3

Convert to LlamaIndex tools
Call mcp_tool_spec.to_tool_list_async() to convert all NVIDIA Vision MCP tools into native FunctionTool objects that any LlamaIndex agent can use.
4

Run with any LLM
Create a FunctionAgent with the tools and your preferred LLM. Swap OpenAI for Anthropic, Gemini, or any LlamaIndex-supported provider.

agent.py

from llama_index.tools.mcp import BasicMCPClient, McpToolSpec
from llama_index.core.agent.workflow import FunctionAgent
from llama_index.llms.openai import OpenAI

# Connect to the MCP
mcp_client = BasicMCPClient(
    "https://edge.vinkius.com/[YOUR_TOKEN_HERE]/mcp"
)
mcp_tool_spec = McpToolSpec(client=mcp_client)

# Convert MCP tools to LlamaIndex tools
tools = await mcp_tool_spec.to_tool_list_async()

# Create and run the agent
agent = FunctionAgent(
    tools=tools,
    llm=OpenAI(model="gpt-4o"),
    system_prompt="You have access to NVIDIA Vision tools.",
)
response = await agent.run("List recent NVIDIA Vision data")

Get your connection token →

Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by NVIDIA. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.

Why Choose Vinkius

Vinkius connects your tools to AI with real-time monitoring and automatic cost savings — all from one dashboard.

Connect NVIDIA Vision now

Real-time monitoring

Live

visibility into every interaction

Connect your favorite tools to your AI and see exactly what's happening — every request, every response, in real time.

Built-in savings

60%

lower AI costs

Vinkius compresses data between your apps and your AI automatically. Lower bills every month — no configuration required.

Single dashboard

One

place for every integration

Every tool your AI connects to, managed from a single screen. One account, complete control.

Common questions about NVIDIA Vision MCP in LlamaIndex

Look, here's the thing: LlamaIndex takes the text outputs from tools like `image_captioning` or `document_qa` and converts them into document nodes. These nodes are then embedded and stored in your configured vector database for future semantic queries.

Yes. When initializing the MCP tool specification, you can pass an allowed_tools list to restrict the agent to specific functions like `detect_objects` while blocking generation tools.

You pass the public image URL directly to the `visual_question_answering` tool within your agent's query loop. The tool returns the answer, which LlamaIndex can then use to synthesize a final response to the user.

You install the MCP tool package, initialize the BasicMCPClient with your Vinkius transport URL, and convert the server tools using McpToolSpec. Pass these tools to your LlamaIndex FunctionAgent to start querying.

Yes, because Vinkius runs the connector in an isolated, ephemeral sandbox. Your raw image files and scanned documents are processed in memory, sent to NVIDIA's secure endpoints, and never written to persistent storage on the platform.

Use it with your favorite AI tools

Connect this server to Cursor, Claude, VS Code, and more.

OpenAI Agents SDK sdk-python

Google ADK sdk-python

Pydantic AI sdk-python

Vercel AI SDK sdk-typescript