Qdrant

Store and retrieve agent output in a Qdrant collection through the Qdrant MCP server.

qdrant.py
"""
Qdrant
=============================

Demonstrates qdrant.
"""

import asyncio
from os import getenv

from agno.agent import Agent
from agno.models.google import Gemini
from agno.tools.mcp import MCPTools
from agno.utils.pprint import apprint_run_response

# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------


QDRANT_URL = getenv("QDRANT_URL")
QDRANT_API_KEY = getenv("QDRANT_API_KEY")
COLLECTION_NAME = "qdrant_collection"
EMBEDDING_MODEL = "sentence-transformers/all-MiniLM-L6-v2"


async def run_agent(message: str) -> None:
    async with MCPTools(
        "uvx mcp-server-qdrant",
        env={
            "QDRANT_URL": QDRANT_URL,
            "QDRANT_API_KEY": QDRANT_API_KEY,
            "COLLECTION_NAME": COLLECTION_NAME,
            "EMBEDDING_MODEL": EMBEDDING_MODEL,
        },
    ) as mcp_tools:
        agent = Agent(
            model=Gemini(id="gemini-2.5-flash-preview-05-20"),
            tools=[mcp_tools],
            instructions="""
            You are the storage agent for the Model Context Protocol (MCP) server.
            You need to save the files in the vector database and answer the user's questions.
            You can use the following tools:
            - qdrant-store: Store data/output in the Qdrant vector database.
            - qdrant-find: Retrieve data/output from the Qdrant vector database.
            """,
            markdown=True,
        )

        response = await agent.arun(message, stream=True)
        await apprint_run_response(response)


# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------

if __name__ == "__main__":
    query = """
    Tell me about the extinction event of dinosaurs in detail. Include all possible theories and evidence. Store the result in the vector database.
    """
    asyncio.run(run_agent(query))

Run the Example

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate

Install dependencies

uv pip install -U "agno[mcp]" google-genai

Prepare uvx

Install uv, then verify uvx is available:

uvx --version

Export environment variables

export GOOGLE_API_KEY="your_google_api_key_here"
export QDRANT_API_KEY="your_qdrant_api_key_here"
export QDRANT_URL="your_qdrant_url_here"

Use an active Gemini model

Save the source above as qdrant.py. Its preview model was shut down. Replace the model declaration in the saved file with:

model=Gemini(id="gemini-3.6-flash"),

Run the example

Run the adapted qdrant.py:

python qdrant.py

Full source: cookbook/91_tools/mcp/qdrant.py