YOU CAN REPLICATE THIS PROJECT -> https://github.com/Nagharjun17/MCP-Ollama-Client

• Runs entirely offline on a local LLM via Ollama (default: qwen3:14b, but any function‑calling model works).
• Connects to multiple MCP servers at once (e.g., postgres, filesystem) declared in a single config.json.
• On startup: launches servers → fetches tool schemas → prefixes tools as <server>.<tool> → hands the merged set to the LLM.
• The model then decides which server/tool to call per user request.
• Everything is configurable in one file; add/remove servers without editing code.

Requirements (tested)

• Python ≥ 3.12
• Ollama ≥ 0.8.0
• MCP servers with stdio transport (e.g., postgres-mcp, server‑filesystem)

Quick start

# 1) clone
git clone https://github.com/Nagharjun17/MCP-Ollama-Client.git
cd mcp-ollama-client

# 2) env
uv venv && source .venv/bin/activate
uv pip sync pyproject.toml    # or: uv pip install -r uv.lock

# 3) local model
ollama pull qwen3:14b

# 4) config + servers (example)
uv pip install postgres-mcp    # and set DATABASE_URI etc. in config.json

# 5) run
uv run client.py

Example boot:

🔌 Starting MCP server: postgres
🔌 Starting MCP server: filesystem
🛠️  Aggregated tools: ['postgres.list_schemas', 'filesystem.read_file', ...]
>>>
MCP Client Startup

Type natural language queries; the LLM will call tools as needed.

config.json (schema excerpt)

{
  "llm": {
    "model": "qwen3:14b",
    "temperature": 0.7,
    "max_tokens": 2048
  },
  "mcpServers": {
    "postgres": {
      "command": "postgres-mcp",
      "args": ["--access-mode=restricted"],
      "env": {
        "DATABASE_URI": "postgresql://user:pass@localhost:5432/db"
      }
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/mnt/smbshare/"]
    }
  }
}

• Keys under mcpServers become prefixes (postgres.*, filesystem.*).
• Each server is launched as its own stdio subprocess.
• Local‑first: no cloud keys required.

GitHub Repo