Dicklesworthstone

Vibe Cockpit — AI skill for Claude Code

AI community

Real-time monitoring dashboard for AI coding agent fleets: session health, output streaming, and observability across Claude, Codex, and Gemini.

How to install Vibe Cockpit

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Dicklesworthstone/vibe_cockpit and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Vibe Cockpit does

Real-time monitoring dashboard for AI coding agent fleets: session health, output streaming, and observability across Claude, Codex, and Gemini.

Alternatives in AI

Manifest — Real-time cost observability for OpenClaw agents — track tokens, costs, messages, and model usage 4.1k ★
Botmux — Bridge Feishu/Lark to AI coding CLIs — Claude Code, Codex, Gemini, OpenCode… every DM, group or topic spawns i 1.2k ★
Tokentap — Intercept LLM API traffic and visualize token usage in a real-time terminal dashboard 763 ★

README

Vibe Cockpit (`vc`)

![Platform](https://img.shields.io/badge/platform-Linux%20%7C%20macOS-blue.svg) ![Rust](https://img.shields.io/badge/rust-2024%20edition-000000.svg) ![License](https://img.shields.io/badge/License-MIT%2BOpenAI%2FAnthropic%20Rider-blue.svg)

Vibe Cockpit is a fleet console for AI coding agents. It collects telemetry from the tools you already run — `ntm`, `caut`, `cass`, `caam`, `dcg`, `br`/`bv`, `ru`, `rano`, `pt`, `rch`, MCP Agent Mail — across local and SSH-reachable machines, lands it in one DuckDB store, scores it, and serves it back as a TUI, a read-only web API, an MCP server, and a JSON CLI that agents can drive.

curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/vibe_cockpit/main/install.sh" | bash

TL;DR

The Problem

Running a dozen agents across a few boxes means a dozen sources of truth. Which account is about to hit a rate limit? Which machine is 3% from a full disk? Which repo has been dirty for six hours? Every one of those answers lives in a different tool, on a different host, behind a different flag.

The Solution

One collector framework, one store, one set of queries. `vc daemon` polls each tool on a schedule, writes the rows, scores every machine, and raises alerts. Everything else — TUI, web, MCP, robot JSON — is a view over that store.

Why It's Different

Agent-first. Every read surface has a machine-readable form. vc --format json and vc robot emit a versioned envelope; there is an MCP server so an agent can ask about the fleet without shelling out at all.
It tells you when it doesn't know. A screen with no backing query says NO DATA SOURCE YET and names the table it would need. It does not invent a number.
Cancel-correct. Built on Asupersync; every collector takes a capability context, so a SIGTERM mid-tick unwinds cleanly instea