VibePod CLI
The vp run command starts Claude Code, Codex, Gemini and other CLI agents in Docker or Podman containers
- 190 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Agents run in containers with access to allowed project folders and to your model credentials
- The `--ikwid` flag turns off approval prompts inside the agent, which then runs commands without asking
Install
- Works in
Manual install
pip install vibepodDocker or Podman must be running. First launch: `vp run <agent>`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agents run in containers with access to allowed project folders and to your model credentials
- The `--ikwid` flag turns off approval prompts inside the agent, which then runs commands without asking
This is third-party code. Review the repository files before installing.
What it does
VibePod is the `vp` CLI that runs coding agents in isolated Docker or Podman containers. Setup is optional: `vp run claude`, `vp run codex` or `vp run vibe` start a container from a ready image in the `vibepod` namespace on Docker Hub. Agents include Claude Code, Gemini CLI, Codex, Mistral Vibe, Copilot, OpenCode, Hermes, Qwen and more. Extra arguments reach the agent after `--`. While an agent runs, VibePod collects metrics on your own machine: HTTP traffic per agent, usage over time and Claude tokens. The dashboard starts with `vp logs start` and lets you compare agents side by side. Nothing goes to the cloud. Other commands handle the rest. `vp provider add` registers an OpenAI- or Anthropic-compatible endpoint or a local model server, `vp skills add` installs skills per project or per user, and a file in `.vibepod/overlay/` extends the agent image for a project. The `--acp` flag turns `vp` into an Agent Client Protocol adapter for Zed and other editors.
Who it is for. Developers who run several AI agents on one machine and want each in its own container with a shared metrics dashboard.
Good fit when
- An agent goes into someone else's repository and runs in a container so it cannot touch the rest of the system
- Claude Code and Codex are compared on one task by traffic, tokens and time in a local dashboard
- A model runs on a local server or a compatible API and gets wired to the agents with one `vp provider add`
- Zed connects a containerized agent through `vp run claude --acp` instead of installing the agent on the host
- The agent image needs a project's toolset, and a Dockerfile fragment is layered on top of the base image
Not a fit when
- The machine has no Docker or Podman and cannot get them
- The agent must work with the host operating system itself: inside a container it sees only the allowed project folder
- You want a single agent without metrics or containers: running it directly is simpler
- You need a network gateway or one policy for a whole organization: VibePod runs on one computer
Example request
Run Codex in a container through VibePod in this repository and start the metrics dashboardLimitations
- A running Docker or Podman is required. Agent images come from Docker Hub, and any image can be replaced with a variable like `VP_IMAGE_CLAUDE`.
- The `--ikwid` flag appends each agent's own approval-skip option, for example `--dangerously-skip-permissions` for Claude Code. OpenCode, Auggie and several other agents have no such option.
- When an editor launches it with `--acp`, standard input is a pipe, so the folder must be allowed in advance with `vp config allow-dir`.
- Token metrics exist for Claude; for other agents the dashboard shows traffic and usage.
How to disable. Stop an agent with `vp stop <agent>` or `vp stop --all`, and the dashboard with `vp logs stop`. Remove the package with the manager you used: `pip uninstall vibepod`, `brew uninstall vibepod` or `conda remove vibepod`.
FAQ
How do I run Claude Code in a container with VibePod?
Install the package with `pip install vibepod`, make sure Docker or Podman is running and run `vp run claude` in the project folder. Pass Claude Code's own arguments after `--`, for example `vp run claude -- --help`.
What does the --ikwid flag do?
It appends the agent's own auto-approval option to the launch: `--dangerously-skip-permissions` for Claude Code, `--approval-mode=yolo` for Gemini, `--dangerously-bypass-approvals-and-sandbox` for Codex. The container remains the boundary, but prompts inside the agent are off.
Where do agent metrics and traffic go?
Nowhere outside your computer. VibePod collects them locally and shows them in the dashboard started by `vp logs start`. Stop it with `vp logs stop`.
Can I plug in my own model?
Yes, `vp provider add` registers any OpenAI- or Anthropic-compatible endpoint or a local model server. You then route an agent to it with the `vp run --provider` flag.
How do I connect VibePod to the Zed editor?
Add an agent server to `settings.json` with the command `vp` and arguments `run claude --acp`. Allow the project folder in advance with `vp config allow-dir`, otherwise the interactive prompt cannot run in the editor.
Discussion
No questions or reviews yet.
Related
CLI toolsA Docker sandbox for coding agents: a Debian image with Copilot CLI, Codex and Pi, and SSH, RDP and Docker enabled by flags.
AgentBox runs Claude Code, Codex and Open Code in separate virtual machines on your PC, a self-hosted server or a cloud.
Container by Aerovato
Container
A CLI for persistent per-project Linux workspaces in Docker or Podman that keep the agent away from the rest of your system
Containarium: isolated boxes for AI agents
Containarium
A platform on your own server that gives each agent its own Linux box with SSH and MCP tools for commands and files
A self-hosted control center for coding agents: Claude Code, Codex, Copilot, OpenCode and Pi from your phone, desktop and terminal
Runtm runs Claude Code, Codex and other agents in isolated sandboxes and deploys the app they build to a live HTTPS endpoint
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| VibePod (this card) | no | 190 | ||
| VibeKit | no | 1,867 | a proxy that redacts secrets, with Docker behind a flag | |
| container-use | yes | not required | 4,084 | Dagger's MCP server giving each agent a container and a git branch |
README in short
The page lists features: zero-config launch, container isolation, one interface for a dozen agents, providers, skills, project overlays and a local dashboard. It continues with installation through pip, Homebrew and conda-forge, the `--ikwid` option table, editor integration over ACP and the image list with overrides.