Ollama Hive MCP Server

What is Ollama Hive MCP Server?

A TypeScript-based Model Context Protocol (MCP) server that integrates with Ollama models using LangChain. It provides tools and resources for querying local language models through a standardized MCP interface, with model-specific MCP server configurations.

How to use Ollama Hive MCP Server?

Run directly via npx: npx ollama-hive-mcp-server. Optionally set MCP_CONFIG_PATH and environment overrides. For persistent installation, clone the repository, install dependencies, build, and start with npm start. The server exposes tools like query_model, test_model, get_model_info, and list_models.

Key features of Ollama Hive MCP Server

• Model management: configure multiple Ollama models with per-model endpoints and settings.
• Model-specific MCP servers: each model can have its own set of MCP tools.
• Session management: conversation sessions for context continuity across queries.
• Environment variable support: override endpoints and configuration via environment variables.
• Stdio transport: standard input/output communication for seamless integration.
• Pre-loading: all models loaded at startup for quick responses.

Use cases of Ollama Hive MCP Server

• Query local Ollama language models from any MCP client (e.g., Claude Desktop, Cursor IDE).
• Build multi-turn conversations with session-based context retention.
• Provide specialized tool sets per model (e.g., filesystem for code model, web search for research model).
• Test and debug model connectivity and MCP server configurations.

FAQ from Ollama Hive MCP Server

What runtime does it require?

Node.js 18+ and an Ollama instance running locally (default http://localhost:11434).

How are model endpoints resolved?

Priority: 1) OLLAMA_ENDPOINT environment variable, 2) model-specific endpoint config, 3) global globalEndpoint config, 4) default http://localhost:11434.

How do I use sessions for context continuity?

Create a session with the create_session tool, then pass the returned sessionId to subsequent query_model calls. Alternatively, set createSession: true to auto-create a session.

Are MCP servers isolated per model?

Yes. Each model can define its own set of MCP servers; they are isolated from other models.

How is configuration validated?

All configuration is validated using Zod schemas; errors are returned in a standardized format.