VibePod

VibePod is a unified CLI (vp) for running AI coding agents in isolated Docker or Podman containers — no required configuration, no setup. Just vp run <agent>. Includes built-in local metrics collection, HTTP traffic tracking, and an analytics dashboard to monitor and compare agents side-by-side.

Features

• ⚡ Zero config — no setup required; vp run <agent> just works. Optional YAML for custom configuration
• 🐳 Isolated agents — each agent runs in its own Docker or Podman container
• 🔀 Unified interface — one CLI for Claude, Gemini, Codex, Devstral/Vibe, Copilot, Auggie, Pi, Agy, Tau, Jcode, Freebuff, Qwen, dsh & more
• 🧩 Skills — install reusable prompt recipes per-project or per-user with vp skills add
• 🧱 Project overlays — commit a FROM-less Dockerfile fragment in .vibepod/overlay/ and VibePod auto-builds a cached, content-addressed image layer on top of the agent's base image — one clearly named image per project and agent (docs)
• 📊 Local analytics dashboard — track usage and HTTP traffic per agent, plus token metrics
• 🐑 Herdr aware — vp run inside a herdr pane reports agent state automatically
• ⚖️ Agent comparison — benchmark multiple agents against each other in the dashboard
• 🔒 Privacy-first — all metrics collected and stored locally, never sent to the cloud
• 📦 Simple install — via pip, Homebrew, or conda-forge

Installation

VibePod is available on PyPI:

pip install vibepod

with Homebrew:

brew install vibepod/vibepod/vibepod

and on conda-forge for conda, mamba, and pixi users:

conda install -c conda-forge vibepod
mamba install -c conda-forge vibepod
pixi global install vibepod

Quick Start

vp run <agent>
# examples:
vp run claude
vp run codex
vp run vibe   # alias of devstral

Extra arguments after the agent are forwarded to the agent process. Use -- before agent flags so VibePod does not parse them as its own options:

vp run <agent> -- <agent-args>

IKWID Mode (--ikwid)

Use --ikwid to append each agent's auto-approval / permission-skip flag when supported.

Agent	--ikwid appended args
claude	--dangerously-skip-permissions
gemini	--approval-mode=yolo
devstral (vibe)	--auto-approve
copilot	--yolo
codex	--dangerously-bypass-approvals-and-sandbox
pi	--approve
agy	--dangerously-skip-permissions
opencode	Not supported
auggie	Not supported
tau	Not supported
jcode	Not supported
freebuff	Not supported
qwen	--approval-mode=yolo
dsh	Not supported

Editor integration (--acp)

vp run <agent> --acp turns VibePod into an Agent Client Protocol adapter, so the containerized agent appears directly in the AI panel of any editor with ACP support (e.g. Zed) — with isolation, profiles, overlays and proxy metrics intact. Supported out of the box: claude, gemini, qwen, codex, opencode, copilot, auggie, jcode, devstral and pi.

Register vp as a custom/external agent server in your editor. Zed example (settings.json):

{
  "agent_servers": {
    "VibePod Claude": {
      "type": "custom",
      "command": "vp",
      "args": ["run", "claude", "--acp"],
      "env": {}
    }
  }
}

Run vp config allow-dir /path/to/project once first (the editor's stdin is a pipe, so the interactive prompt cannot run). See the ACP docs for details and limitations.

Tool Thumbnails

Current Status

This repository contains an initial v1 implementation with:

• vp run <agent>
• vp stop <agent|--all>
• vp list
• vp config init
• vp config show
• vp config path
• vp version

Analytics & Dashboard

VibePod collects metrics locally while your agents run and serves them through a built-in dashboard.

Command	Description
vp logs start	Start or resume dashboard for collected metrics
vp logs stop	Stop the dashboard container
vp logs status	Show dashboard container status

The dashboard shows per-agent HTTP traffic, usage over time, and Claude token metrics. It also lets you compare agents side-by-side. All data stays on your machine.

Image Namespace

All agent images are published under the vibepod namespace on Docker Hub. Source Dockerfiles are in VibePod/vibepod-agents.

Current defaults:

• claude -> vibepod/claude:latest
• gemini -> vibepod/gemini:latest
• opencode -> vibepod/opencode:latest
• devstral (alias: vibe) -> vibepod/devstral:latest
• auggie -> vibepod/auggie:latest
• copilot -> vibepod/copilot:latest
• codex -> vibepod/codex:latest
• pi -> vibepod/pi:latest
• agy -> vibepod/agy:latest
• tau -> vibepod/tau:latest
• jcode -> vibepod/jcode:latest
• freebuff -> vibepod/freebuff:latest
• qwen -> vibepod/qwen:latest
• dsh -> vibepod/dsh:latest
• datasette -> vibepod/datasette:latest
• proxy -> vibepod/proxy:latest (repo)

Overriding Images

You can override any single image directly:

VP_IMAGE_CLAUDE=vibepod/claude:latest vp run claude
VP_IMAGE_GEMINI=vibepod/gemini:latest vp run gemini
VP_IMAGE_OPENCODE=vibepod/opencode:latest vp run opencode
VP_IMAGE_DEVSTRAL=vibepod/devstral:latest vp run devstral
VP_IMAGE_DEVSTRAL=vibepod/devstral:latest vp run vibe   # same agent/image as devstral
VP_IMAGE_AUGGIE=vibepod/auggie:latest vp run auggie
VP_IMAGE_COPILOT=vibepod/copilot:latest vp run copilot
VP_IMAGE_CODEX=vibepod/codex:latest vp run codex
VP_IMAGE_PI=vibepod/pi:latest vp run pi
VP_IMAGE_AGY=vibepod/agy:latest vp run agy
VP_IMAGE_TAU=vibepod/tau:latest vp run tau
VP_IMAGE_JCODE=vibepod/jcode:latest vp run jcode
VP_IMAGE_FREEBUFF=vibepod/freebuff:latest vp run freebuff
VP_IMAGE_QWEN=vibepod/qwen:latest vp run qwen
VP_IMAGE_DSH=vibepod/dsh:latest vp run dsh
VP_DATASETTE_IMAGE=vibepod/datasette:latest vp logs start
VP_SKILLS_ENGINE_IMAGE=vibepod/skills-engine:latest vp skills list

License

MIT License - see LICENSE for details.