Top AI & Machine Learning MCP servers

Ranked by the composite AgentRank score — a weighted blend of stars (15%), freshness (25%), issue health (25%), contributors (10%), and inbound dependents (25%). Average score across these 8 tools: 89.0.

See all ai-models tools in the index →

Choosing by use case

Local model inference (Ollama)

Run Llama, Mistral, and other open models locally and call them from your MCP-enabled agent.

Vector databases

Store and retrieve embeddings for RAG, semantic search, and memory — Pinecone, Qdrant, Chroma, Weaviate.

LLM API gateways

Route between OpenAI, Anthropic, Gemini, and open models through a unified MCP interface.

Embedding generation

Generate text embeddings on demand for clustering, similarity search, and knowledge base construction.

Quick setup

Most MCP servers follow the same config pattern. Add this to your Claude Desktop, Cursor, or Windsurf MCP config file:

{
  "mcpServers": {
    "ollama": {
      "command": "npx",
      "args": ["-y", "ollama-mcp"]
    }
  }
}

See the integrations page for platform-specific setup guides for Claude Code, Cursor, VS Code, Windsurf, and Cline.

Reading the AgentRank signals

Each tool in this guide is scored on five signals that predict long-term reliability:

Full scoring methodology →