MCP Server Ollama Bridge

Use local Ollama models from any MCP-compatible AI client.

Part of the HumoticaOS "MCP for Any AI" initiative - no vendor lock-in, your choice of AI, same powerful tools.

Why?

MCP (Model Context Protocol) lets AI assistants use external tools. But what if you want to use Ollama's local models as part of your workflow? This bridge exposes Ollama to any MCP client.

Use cases:

• Let Claude Desktop query your local Qwen/Llama for sensitive data
• Have Gemini use your local models for embeddings
• Run hybrid workflows: cloud AI for creativity, local for privacy

Installation

pip install mcp-server-ollama-bridge

Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "ollama": {
      "command": "python3",
      "args": ["-m", "mcp_server_ollama_bridge"],
      "env": {
        "OLLAMA_URL": "http://localhost:11434"
      }
    }
  }
}

Available Tools

Examples

Chat:

ollama_chat(message="Explain quantum computing", model="qwen2.5:7b")

Code generation:

ollama_generate(prompt="Write a Python function to sort a list", model="qwen2.5:32b")

Embeddings for RAG:

ollama_embeddings(text="HumoticaOS is an AI orchestration platform")

Part of HumoticaOS

This is one of our published MCP servers:

• mcp-server-rabel - AI Memory & Communication
• mcp-server-tibet - Trust & Provenance
• mcp-server-inject-bender - Security Through Absurdity
• mcp-server-ollama-bridge - Local LLM Bridge

License

MIT - By Claude & Jasper from HumoticaOS, Kerst 2025

One love, one fAmIly