add-ollama-tool

Exposing local Ollama models via an MCP server enables the container agent to call local models. This integration provides tools for model discovery and generation, allowing containerized workflows to access local models for fast, offline-friendly inference.

Yes, using local models for offline inference inside a containerized workflow is possible by exposing an Ollama MCP server. This setup allows the container agent to perform low-latency tasks like summarization or code generation without relying on external network calls.

Before calling Ollama models from your container agent, you need Ollama installed and running on the host machine. You must also pull at least one local model and restart the container to activate the local model calls.

An Ollama MCP server handles model discovery and generation by providing two specific tools. The ollama_list_models tool discovers available local models, and the ollama_generate tool produces responses, enabling direct model interaction within the containerized environment.

Use local Ollama models instead of external API calls when containerized workflows require fast, offline-friendly, or low-latency inference. This approach is ideal for localized tasks such as summarization or code generation where minimizing network dependency is critical.