ollama-local
Manage and use local Ollama models. Use for model management (list/pull/remove), chat/completions, embeddings, and tool-use with local LLMs. Covers OpenClaw sub-agent integration and model selection guidance.
Why use this skill?
Learn to manage and run local LLMs with OpenClaw. Integrate privacy-focused AI models for coding, reasoning, and multi-agent tasks using Ollama.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/timverhoogt/ollama-localWhat This Skill Does
The ollama-local skill for OpenClaw provides a powerful interface to manage and interact with local Large Language Models (LLMs) via the Ollama ecosystem. It bridges the gap between your local infrastructure and the agentic capabilities of OpenClaw, allowing you to pull, run, list, and delete models directly from the command line or within your automation scripts. By leveraging local models, you gain the ability to process data, write code, and reason about tasks without sending sensitive information to external cloud APIs, ensuring maximum privacy and zero latency costs.
Installation
To integrate this capability into your environment, use the OpenClaw hub command. Ensure you have the Ollama server installed and running on your host machine (default port 11434). Run the following command in your terminal:
clawhub install openclaw/skills/skills/timverhoogt/ollama-local
Once installed, verify your connectivity by checking the OLLAMA_HOST environment variable, which defines whether your agent connects to a local daemon or a remote Ollama server instance.
Use Cases
This skill is highly versatile for developers and power users. Common applications include:
- Privacy-First Data Analysis: Run local models to summarize documents or extract insights from sensitive files without data leaving your local network.
- Automated Coding Pipelines: Use dedicated coding models like qwen2.5-coder to perform automated code reviews, refactoring, or snippet generation as part of a CI/CD process.
- Agentic Multi-Tasking: Utilize the 'Think Tank' pattern by spawning parallel sub-agents (e.g., an architect, a coder, and a reviewer) to solve complex software problems collaboratively.
- Tool Orchestration: Enable local models to call external tools, such as performing web searches or interacting with local system utilities through the ollama_tools wrapper.
Example Prompts
- "Spawn a new session using the llama3.1:8b model to analyze my README.md file and suggest improvements for clarity."
- "Use the ollama-local skill to list all currently installed models and then pull the latest deepseek-r1:8b checkpoint."
- "Set up a multi-agent loop with qwen2.5-coder as the primary developer and qwen3:4b as the code reviewer to debug the main.py script."
Tips & Limitations
To get the most out of ollama-local, focus on model selection based on your hardware constraints. Use smaller models like qwen3:4b for rapid prototyping or simple tasks, and reserve larger 12b+ models for complex reasoning. Always verify tool-use support, as not all models natively support the function-calling format required for advanced tool loops. Since this skill runs locally, the performance is entirely dependent on your system's GPU/VRAM capacity. If you encounter slow responses, ensure you have sufficient memory allocation for your chosen quantization level.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-timverhoogt-ollama-local": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: network-access, code-execution