ollama-cli

name	ollama-cli
description	Execute and manage local LLMs using the ollama CLI, including pulling models and launching agents with MCP configurations. You MUST load this skill when interacting with the `ollama` CLI.
license	MIT

ollama-cli

When to Use

• When querying, pulling, or managing local LLMs via the ollama CLI.
• When configuring local model backends for AI agents.
• When launching wrapper tools (like Claude Code) using an Ollama model with Model Context Protocol (MCP) configurations.

When Not to Use

• When making standard REST API calls to remote cloud providers (e.g., OpenAI, Anthropic) without a local proxy.
• For managing Python environments or general CLI tools unrelated to local inference.

Step-by-Step Workflows

1. Verification and Setup

1. Check Daemon Status: Run ollama list to verify the local Ollama daemon is active and responsive.
2. Pull Model: If the target model is missing, execute ollama pull <model-name> (e.g., ollama pull gemma4:e2b) to cache it locally before running.

2. Execution and Agent Launch

1. Interactive Run: Use ollama run <model-name> to test the model directly in the CLI.
2. Launch with MCP Config: When integrating Ollama models with agent workflows (like Claude Code) that require Model Context Protocol configurations, use process substitution to pass dynamic JSON configurations safely.

Command Examples