Clawrium CLI
The clawctl CLI manages a fleet of AI agents on local machines over SSH and Ansible: install, start, upgrade and token tracking.
- 67 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The client creates a passwordless sudo user on hosts and manages agents that have access to the machines.
- The control machine stores provider keys and secrets for the fleet.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/clawriumDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The client creates a passwordless sudo user on hosts and manages agents that have access to the machines.
- The control machine stores provider keys and secrets for the fleet.
This is third-party code. Review the repository files before installing.
What it does
Clawrium installs as a `uv` tool and provides the `clawctl` command line for an agent fleet. Hosts register over SSH with a service user `xclm` on each, after which `clawctl agent create` deploys an agent of the chosen type: OpenClaw, Hermes, ZeroClaw or Claude Code. Ansible does the work underneath, with no containers or Kubernetes needed. Inference providers (Anthropic, OpenAI, OpenRouter, Ollama) are registered separately and assigned to agents. The client tracks token spend, sets limits, swaps models on agents without editing configs, and handles upgrades, rollbacks, secrets rotation and backups. For an overview there is the `clawctl gui` web panel: agents, topology, providers, skills and integrations.
Who it is for. Engineers of home labs and small teams who run several AI agents on several machines and lose track of configs and spend.
Good fit when
- Agents of different types run on three machines and one shared status list is needed.
- Several agents need one local model on a Mac mini or a server.
- Models must be compared across agents without editing each config.
- A token limit is needed before an experiment so the API budget is not burned.
Not a fit when
- There is a single agent on one machine.
- A hosted dashboard with accounts and subscriptions is required.
- Hosts cannot have a user with passwordless sudo.
Example request
Register host worker-1, deploy an OpenClaw agent on it with the Anthropic provider and show fleet status with `clawctl agent get`.Limitations
Each host needs an `xclm` user with passwordless `sudo` and an SSH key: `clawctl host create` prints the setup command and you paste it on the host by hand. Ubuntu and macOS are supported (enable Remote Login on macOS), Discord and Slack channels work for OpenClaw only, and Claude Code installs without a daemon and runs on demand. Python 3.10, `uv` and a provider key are required. It is a client for a managed fleet and not a hosted service: backups and keys live on the control machine.
How to disable. Stop and remove agents with `clawctl agent`, remove hosts through `clawctl host`, then uninstall the tool with `uv tool uninstall clawrium`. Delete the `xclm` user on hosts manually.
FAQ
Which agents are supported?
OpenClaw, Hermes and ZeroClaw with a daemon, plus Claude Code as a daemonless command environment. IronClaw is planned.
Are containers required?
No. The fleet is managed over SSH through Ansible, with no Docker or Kubernetes.
Which inference providers exist?
Anthropic, OpenAI, OpenRouter and Ollama. Claude Code takes an Anthropic API key or an explicitly supplied OAuth token.
How do I see fleet state?
`clawctl agent get` prints name, type, host, provider and status. The web panel opens with `clawctl gui`.
Is it a cloud service?
No. There is no sign-up or cloud dashboard: a Python CLI talks to your machines through Ansible.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI agent swarm: a leader spawns workers in separate git worktrees, and they share tasks and exchange messages via files.
ClawMetry reads the session files agents already write and shows the timeline, tool calls, tokens and cost in one dashboard.
One runtime for Claude Code, Codex, OpenCode and more: a run counts as done only when checks the agent did not write pass
A CLI that connects local agents to a shared cloud store of sessions, memory, skills and MCP connectors.
AgentOS runs a team of OpenClaw agents: workspaces, scheduled tasks, activity reports and human approvals
A compact OpenClaw alternative: a personal assistant on Claude Agent SDK with agents in isolated containers and messengers added via skills
README in short
The client is built like `kubectl` for agents: one command set for all hosts. The README lists agent types, supported OSes and providers and a five-step quickstart.