← Back to Skills Marketplace
Ollama Manager
by
Twin Geeks
· GitHub ↗
· v1.3.1
· MIT-0
257
Downloads
0
Stars
3
Active Installs
6
Versions
Install in OpenClaw
/install ollama-manager
Description
Manage Ollama models across your machines — see what's loaded, what's eating disk, what's never used, and what you should pull next. Get AI-powered recommend...
Usage Guidance
This skill appears coherent for managing Ollama models, but it recommends installing and running a third-party PyPI package and running a long‑running 'herd' router which will read local telemetry and control pulls/deletes. Before installing or running it: review the ollama-herd PyPI package and its GitHub repo for provenance and code behavior; run installation and the herd-node in an isolated/test environment (or container); confirm the router's network binding and authentication (ensure it doesn't bind to 0.0.0.0 without auth); inspect ~/.fleet-manager/latency.db and logs if they contain sensitive data; and avoid running installs as root. If you cannot review the package, treat the pip install step as higher risk.
Capability Assessment
Purpose & Capability
The name/description (manage Ollama models across machines) matches the instructions: install the ollama-herd toolkit, run a herd router and herd-node, query router endpoints for model lists/disk usage, and query a local SQLite telemetry DB for usage/latency. The metadata's listed bins and configPaths (~/.fleet-manager/latency.db, ~/.fleet-manager/logs/herd.jsonl) are directly relevant to that purpose.
Instruction Scope
The SKILL.md tells the agent to: pip install a third-party package, run herd/herd-node daemons, curl localhost:11435 endpoints, and read/query ~/.fleet-manager/latency.db and logs. Those actions are within scope for fleet management, but they do give the skill access to local telemetry and allow it to trigger network activity (pull/delete models) via the router. The instructions do not request unrelated files, env vars, or remote endpoints beyond the documented local router and recommended PyPI package.
Install Mechanism
There is no formal install spec in the skill bundle (instruction-only). The SKILL.md recommends installing a PyPI package (pip install ollama-herd). Installing a package from PyPI is expected for this functionality but carries the usual risk that arbitrary code will be installed and run on the machine; this is proportionate to the stated purpose but worth vetting before install.
Credentials
The skill does not request environment variables, credentials, or unrelated config paths. The only declared config paths are local telemetry DB and logs used for usage/latency queries, which are appropriate for the stated diagnostics/recommendation features.
Persistence & Privilege
The skill is instruction-only and not marked always:true. It does not request persistent elevated platform privileges or access to other skills' configurations. Running the recommended herd daemon will create a persistent router process, which is normal for this functionality but is outside the skill bundle itself.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ollama-manager - After installation, invoke the skill by name or use
/ollama-manager - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.1
Cross-platform support: macOS, Linux, and Windows. Updated OS metadata, descriptions, and hardware recommendations.
v1.3.0
- Updated branding and documentation to emphasize "Ollama" throughout the skill.
- Clarified feature descriptions, troubleshooting steps, and commands to reference Ollama models and functionality explicitly.
- Added multilingual summary to description (Chinese and Spanish).
- Slight increment in version number from 1.0.0 to 1.0.1.
- Improved clarity for onboarding and usage instructions by reinforcing "Ollama Manager" naming and terminology.
v1.2.0
- Updated the skill description to explicitly mention support for popular models like Llama, Qwen, DeepSeek, Phi, and Mistral.
- Simplified and clarified the skill's capabilities in the description.
- No code changes; documentation update only.
v1.1.0
- Fixed metadata formatting in SKILL.md.
- Moved the "os" field into "requires" within the metadata block.
- No user-facing feature changes; documentation only.
v1.0.1
- Added optionalBins field to metadata, listing optional dependencies (python3, sqlite3, pip).
- Added configPaths to metadata, indicating relevant config/database/log files.
- No changes to core instructions, usage or documentation.
v1.0.0
Initial release of ollama-manager: centralized Ollama model management across multiple machines.
- View loaded, unused, and disk-heavy models across your fleet.
- Get AI-powered recommendations for optimal model selection based on your hardware.
- Pull, delete, and organize models remotely (no SSH required).
- Track model usage, disk consumption, and performance history with per-machine details.
- Visual dashboard for real-time fleet health and management actions.
Metadata
Frequently Asked Questions
What is Ollama Manager?
Manage Ollama models across your machines — see what's loaded, what's eating disk, what's never used, and what you should pull next. Get AI-powered recommend... It is an AI Agent Skill for Claude Code / OpenClaw, with 257 downloads so far.
How do I install Ollama Manager?
Run "/install ollama-manager" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Ollama Manager free?
Yes, Ollama Manager is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Ollama Manager support?
Ollama Manager is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Ollama Manager?
It is built and maintained by Twin Geeks (@twinsgeeks); the current version is v1.3.1.
More Skills