VibeKit CLI
VibeKit runs Claude Code, Codex and other CLI agents through a proxy that masks secrets in replies, keeps a log and can hold the agent in Docker
- 1,867 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The proxy passes all of the agent's HTTP and HTTPS traffic through itself
- Session logs and statistics pile up on disk in `~/.vibekit` and hold the agent's working record
- With the `--sandbox` flag it starts Docker or Podman containers and mounts host folders
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/vibekitDetects 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 proxy passes all of the agent's HTTP and HTTPS traffic through itself
- Session logs and statistics pile up on disk in `~/.vibekit` and hold the agent's working record
- With the `--sandbox` flag it starts Docker or Podman containers and mounts host folders
This is third-party code. Review the repository files before installing.
What it does
VibeKit is a wrapper for console agents, the `vibekit` package from npm. Instead of `claude` you run `vibekit claude`, instead of `codex` you run `vibekit codex`, and the same for Gemini CLI, Grok CLI, Cursor Agent and OpenCode. The remaining arguments go to the agent unchanged. VibeKit does not store API keys: the agent takes them from its usual environment variables. Settings open with the `vibekit` command and live in `~/.vibekit/settings.json`. A proxy server sits between the agent and its API address. It passes replies through a set of more than 200 patterns, looks for AWS, OpenAI, GitHub and Google keys, database connection strings, email addresses, phone numbers and card numbers, and replaces what it finds with labels like `[OPENAI_API_KEY_REDACTED]`. The rules come from the `rules-stable.yml` file. If loading them fails, only the email and card-number patterns remain. A log of each session is kept in JSON by day and by agent in `~/.vibekit/logs/`, and statistics gather in `~/.vibekit/analytics/`. The `vibekit dashboard` command opens a web panel with sessions and charts, `vibekit logs` and `vibekit analytics` show the same in the terminal, and `vibekit clean` erases what has accumulated. With the `--sandbox` flag the agent works inside a Docker or Podman container, and `vibekit sync` brings the edits from the container back into the project.
Who it is for. For developers who run Claude Code, Codex or Gemini CLI in a terminal and want a session log, secrets masked in replies and, if needed, the agent isolated in a container.
Good fit when
- An agent replies with snippets that carry keys and connection strings, and they must not stay in plain text in your terminal
- You need a log of what the agent did, how long sessions ran, what share finished successfully and how many files changed
- You want to run risky or experimental code in a Docker or Podman container rather than on your work machine
- A team uses different agents and wants one `vibekit dashboard` panel to compare their sessions
- You want to keep the plain `claude` and `gemini` commands: `vibekit setup-aliases` sets up shared aliases
Not a fit when
- You need to stop a dangerous command before it runs: VibeKit masks replies and keeps a log, while checking agent actions is what HOL Guard does
- You need a separate container and git branch for each agent: container-use does that
- You work on Windows without WSL2: the documentation advises running VibeKit inside WSL2
- You need custom redaction patterns, allowlists or reports on matches: the documentation lists them as not yet implemented
Example request
Install VibeKit and run Claude Code through it in this project so that replies with keys are masked and code runs in DockerLimitations
- The sandbox is off by default (`sandbox.enabled` is `false`), and you turn it on with the `--sandbox` flag or `VIBEKIT_SANDBOX=true`. The documentation itself states that it gives process isolation rather than full protection, and mounted host folders weaken that isolation.
- Redaction applies to replies that pass through the proxy and only by patterns. Without the proxy it does not work.
- It needs Node.js 18 or newer and an installed agent, plus Docker or Podman for containers. API keys come from the `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`, `OPENAI_API_KEY` and `GROK_API_KEY` variables.
- Package `vibekit` is at version 0.0.4, and the latest changes were pushed on 13 January 2026.
- The same repository holds the `@vibe-kit/sdk` and `@vibe-kit/auth` packages for embedding into applications, and they are separate from the console wrapper. MIT license.
How to disable. Run the agent directly without `vibekit`, and if you set up aliases, remove them from the shell profile. `npm uninstall -g vibekit` removes the package. The proxy stops with `vibekit proxy kill --port 8080`, and `vibekit sandbox clean` removes containers and images. Logs and settings live in `~/.vibekit`, and `vibekit clean` erases logs and statistics.
FAQ
Which agents run through VibeKit?
Claude Code (`vibekit claude`), Gemini CLI, Codex, Grok CLI, Cursor Agent (`vibekit cursor-agent`) and OpenCode. All of them get the shared `--sandbox`, `--sandbox-type` and `--proxy` options, and `vibekit logs --agent` and `vibekit analytics --agent` filter data by agent.
What does the proxy mask?
The patterns look for AWS keys (including `AKIA...`), OpenAI keys shaped like `sk-...`, GitHub tokens, Google keys and service accounts, database connection strings, email addresses, phone numbers and card numbers. What is found is replaced with a label carrying the pattern name.
How do I choose between Docker and Podman?
`--sandbox-type docker` or `--sandbox-type podman` sets the runtime for one command, and the `VIBEKIT_SANDBOX_TYPE` variable or the `sandbox.type` field in settings makes the choice permanent. `vibekit sandbox build` builds the image and `vibekit sandbox status` shows readiness.
What data does the analytics collect?
The number of sessions, their duration, the share of successful ones, the number of changed files, errors and warnings, and a breakdown by agent. The data sits in `~/.vibekit/analytics/`, and `vibekit analytics --export analytics.json` exports it to JSON.
How do I open the dashboard on another port?
`vibekit dashboard --port 3001` starts the panel on the port you need, and the `--no-open` flag keeps it from opening a browser itself. `vibekit dashboard stop` stops the panel.
Discussion
No questions or reviews yet.
Related
CLI toolsThe safehouse command runs Claude Code, Codex and other agents in a macOS sandbox: files are closed by default and you open folders with flags
A skill that audits vibe-coded apps: secrets, Supabase RLS, authorization, payments, rate limits, mobile and AI integrations
AgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
VibeSec
VibeSec-Skill
A secure coding skill: the agent looks at code from a bug hunter's perspective and prevents common web vulnerabilities
cli sandbox for ai agents without containers or daemons: runs Claude Code, Codex, or OpenCode with restricted access to files, network, and credentials
A CLI with policies that intercepts dangerous AI agent commands like rm -rf, force push or reading .env before they run
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| VibeKit (this entry) | no | 1,867 | ||
| HOL Guard | yes | 971 | checks an agent's actions before they run and holds risky ones until you decide | |
| container-use | yes | not required | 4,078 | gives each agent its own container and git branch for parallel work |
README in short
The repository's front page presents VibeKit as a safety layer for coding agents: the `npm install -g vibekit` install, five properties (local sandbox, redaction, observability, support for different agents, working without the cloud) and links to the SDK and Auth packages. The main description sits in the `docs` folder: pages for each agent, the CLI, authentication and an API reference. The repository holds Dockerfile images for the agents, packages for cloud sandboxes (e2b, daytona, modal and others), a dashboard and CI settings.