DeepSeek + MCP Tools: Open Source AI with External Tools

DeepSeek produces some of the most capable open-source AI models available. By connecting DeepSeek models to MCP (Model Context Protocol) servers, you get a powerful, cost-effective AI system with access to external tools. This guide covers the setup.

What Is DeepSeek?

DeepSeek is a Chinese AI company that produces open-weight LLMs competitive with frontier models. Key models include:

• DeepSeek-V3: General-purpose model with strong reasoning
• DeepSeek-Coder: Optimized for programming tasks
• DeepSeek-R1: Reasoning-focused model with chain-of-thought

Models are available through the DeepSeek API or can be run locally via Ollama.

Step 1: Access DeepSeek Models

Option A: DeepSeek API

pip install openai

# DeepSeek API is OpenAI-compatible
export DEEPSEEK_API_KEY=your_key_here

from openai import OpenAI

client = OpenAI(
    api_key="your_key_here",
    base_url="https://api.deepseek.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

Option B: Local via Ollama

ollama pull deepseek-v2:16b
ollama pull deepseek-coder-v2:16b

Step 2: Install MCP Servers

# Filesystem access
npm install -g @modelcontextprotocol/server-filesystem

# Database access
pip install mcp-server-sqlite

# Web search
npm install -g @modelcontextprotocol/server-brave-search

Step 3: Connect DeepSeek to MCP Tools

Use a framework like LangChain to bridge DeepSeek and MCP:

from langchain_openai import ChatOpenAI
from langchain_mcp_adapters import MCPToolkit
from mcp import StdioServerParameters

# DeepSeek as LangChain LLM (OpenAI-compatible API)
llm = ChatOpenAI(
    model="deepseek-chat",
    api_key="your_key",
    base_url="https://api.deepseek.com/v1"
)

# Connect MCP filesystem server
toolkit = MCPToolkit(
    server_params=StdioServerParameters(
        command="npx",
        args=["-y", "@modelcontextprotocol/server-filesystem", "/workspace"]
    )
)
await toolkit.initialize()
tools = toolkit.get_tools()

# Build agent
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant with file access."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}")
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)
result = executor.invoke({"input": "List Python files in the workspace"})

Step 4: Local Deployment with Full Tool Access

For complete privacy, run DeepSeek locally via Ollama and connect to local MCP servers:

# Start Ollama with DeepSeek
ollama serve &

# All data stays on your machine:
# - Model runs locally
# - MCP servers run locally
# - No external API calls needed

Step 5: Find the Right Tools with XLUXX

pip install xluxx

# Discover MCP servers by capability
curl https://api.xluxx.net/v1/tools?q=code+analysis&sort=trust_score

# Check a specific server
curl https://api.xluxx.net/v1/tools/mcp-server-filesystem

XLUXX trust scores help you choose between competing MCP servers based on security, maintenance, and community validation.

DeepSeek + MCP Use Cases

• Code assistant: DeepSeek-Coder + filesystem MCP for reading and modifying code
• Data analysis: DeepSeek-V3 + SQLite MCP for querying local databases
• Research: DeepSeek-R1 + web search MCP for deep research tasks
• DevOps: DeepSeek-Coder + GitHub MCP for repository management

Cost Comparison

• DeepSeek API: Significantly cheaper than GPT-4o or Claude for similar capability
• Local via Ollama: Zero ongoing cost after hardware investment
• MCP servers: Free and open source

Use XLUXX Trust Layer to find reliable MCP tools: api.xluxx.net

Related Articles

• How to Build an MCP Server (Step-by-Step)
• AutoGen + MCP: Building Reliable Multi-Agent Systems
• Claude Desktop MCP Setup: Install and Configure Tools