claudectl plugin
A Claude Code plugin and CLI: session dashboard, local-LLM review of tool calls, budgets and a brain shared across machines.
- 204 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The hook gate intervenes in every Bash, Write and Edit call and in `--auto-run` mode decides for the user.
- The brain learns from your decisions and shares them with other hive machines.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/claudectlDetects 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 hook gate intervenes in every Bash, Write and Edit call and in `--auto-run` mode decides for the user.
- The brain learns from your decisions and shares them with other hive machines.
This is third-party code. Review the repository files before installing.
What it does
claudectl watches Claude Code sessions. Dashboard mode shows every session, spend, loops and context saturation. A local model, the "brain", learns from your approvals and rejections and allows safe calls itself (reads, searches, test runs) while stopping dangerous ones, such as force pushes or destructive commands, before they run. The plugin from the repository adds a hook gate on Bash, Write, Edit and NotebookEdit calls, the slash commands `/brain`, `/inbox`, `/spend` and `/sessions`, a supervisor sub-agent and a bus MCP server. Several machines can join a "hive" and exchange what the brain has learned. It works with any OpenAI-compatible endpoint: Ollama, llama.cpp, vLLM, LM Studio.
Who it is for. Developers running several Claude Code sessions in parallel who want to see spend and block dangerous commands automatically.
Good fit when
- Parallel sessions burn the budget and you need a spend panel per session.
- An agent looped or lost context and the session must be stopped in time.
- Approval prompts are tedious, but force pushes and deletions must be blocked.
- A second machine should immediately get the first one's settings and experience.
Not a fit when
- You need a predictable rule set with no learning: brain decisions change as data accumulates.
- You have no local model or resources to run one: only a built-in rule set works without it.
- You forbid automatic execution: `--auto-run` lets the brain act without asking.
Example request
Show spend across all my Claude Code sessions this week and turn the brain on in advisory mode.Limitations
The brain decides for you: in `--auto-run` mode it approves and rejects calls without asking, so start in advisory mode. Without a local LLM the hook gate falls back to built-in rules. Building from source needs rustc 1.88 or newer. After a brew upgrade run `claudectl init --upgrade`, otherwise hooks and plugin files go stale. Dashboard numbers come from your own data and promise no savings. The last commit dates to November 2026.
How to disable. Remove the plugin with Claude Code commands, drop the hooks through `claudectl init` and uninstall the binary with `brew uninstall claudectl`.
FAQ
How do I install claudectl?
Through Homebrew: `brew install mercurialsolo/tap/claudectl`, or `cargo install claudectl`. Then run `claudectl init` and check `claudectl doctor`.
How do I add the Claude Code plugin?
Run `claude plugin marketplace add mercurialsolo/claudectl`, then `claude plugin install claudectl@claudectl`.
What does the brain do?
Allows safe calls, blocks dangerous ones, stops looping sessions and routes summaries between sessions. Start in advisory mode: `claudectl --brain`.
Is a local LLM required?
For the brain yes, for example `ollama pull gemma4:e4b`. Without it the hook gate uses built-in rules.
What is a hive?
A named group of machines: it announces itself on the LAN, hands out invites and queues join requests for your approval.
Discussion
No questions or reviews yet.
Related
PluginsA dashboard for every Claude Code session on the machine: live agents, cost, history search and MCP tools for Claude.
claude-mods adds 112 skills, 3 agents and 3 commands to Claude Code: stacks, agent operation, defense, session continuity.
Claudette runs several Claude Code sessions in parallel, each in its own git worktree, from a window or the console
An environment for Claude Code that keeps context across sessions via hooks, ledgers and handoffs, orchestrates agents and saves tokens
A plugin that turns one Claude Code session into a controller handing tasks to workers in tmux and collecting their results.
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
README in short
The README shows a sample impact scorecard, install commands and a five-step quick start. The brain with its learning levels, hive mode and session health monitoring follow.