The verification layer for coding agents.
Vibium gives AI agents the tools they need to check their work. Install the browser skill and your agent can navigate pages, fill forms, click buttons, and take screenshots — all through simple CLI commands. Also available as an MCP server and as JS/TS, Python, and Java client libraries.
New here? Get started in JavaScript, Python, or Java — zero to hello world in 5 minutes.
- AI-native. Install as a skill — your agent learns the full browser automation toolkit instantly.
- Zero config. One install, browser downloads automatically, visible by default.
- Standards-based. Built on WebDriver BiDi, not proprietary protocols controlled by large corporations.
- Lightweight. Single ~10MB binary. No runtime dependencies.
- Flexible. Use as a CLI skill, MCP server, or JS/Python/Java library.
npm install -g vibium
npx skills add https://github.com/VibiumDev/vibiumSelect all Vibium skills in the installer, then choose your agent and installation scope. Vibium installs Chrome automatically.
- browser — explore pages, automate actions, record sessions, and carry out browser tasks from plain-language instructions.
- check — independently assess whether a claim holds in the browser or a recording.
Invoke /browser or /check in agents with slash skills, or $browser and
$check in Codex. Setup guides: Codex,
Claude Code, and
Claude Desktop Chat via MCP.
Run vibium ready to check setup.
For a walkthrough, follow Your coding agent’s first Check.
Run and Check currently require the development build and AI configuration:
run vibium config init, fill in the file it writes at ~/.config/vibium/ai.env,
then source it. See model providers.
See Introducing Run and Check for examples and limitations.
# Give a browser task (requires AI setup)
vibium "open https://var.parts and add one battery pack to the cart" --keep-open
# Or use the explicit run command
vibium run "open https://var.parts and add one battery pack to the cart" --keep-open
# Check the cart
vibium check "the cart contains exactly one battery pack"
# Check and save a recording
vibium check "the cart contains exactly one battery pack" -o cart-check.zip
# Close the browser
vibium stop
# Map & interact (the core workflow)
vibium go https://var.parts # navigate to URL
vibium map # map interactive elements → @e1, @e2, ...
vibium click @e1 # click using ref
vibium diff map # see what changed
# Find elements (semantic — no CSS needed)
vibium find text "Sign In" # find by visible text
vibium find label "Email" # find by form label
vibium find placeholder "Search" # find by placeholder
vibium find role button # find by ARIA role
# Read & capture
vibium text # get all page text
vibium screenshot -o page.png # capture screenshot
vibium screenshot --annotate -o a.png # annotated with element labels
vibium pdf -o page.pdf # save page as PDF
vibium eval "document.title" # run JavaScript
# Wait for things
vibium wait ".modal" # wait for element to appear
vibium wait url "/dashboard" # wait for URL change
vibium wait text "Success" # wait for text on page
# Record sessions
vibium record start # record with screenshots
vibium record stop # stop and save to record.zip
# Forms & input
vibium fill @e2 "hello@example.com" # fill input using ref
vibium select @e3 "US" # pick dropdown option
vibium set @e4 # check a checkbox
vibium press Enter # press a keyFull command list: SKILL.md
Alternative: MCP server (for structured tool use instead of CLI):
claude mcp add vibium -- npx -y vibium mcp # Claude Code
gemini mcp add vibium npx -y vibium mcp # Gemini CLISee MCP setup guide for options and troubleshooting.
npm install vibium # JavaScript/TypeScript
pip install vibium # PythonJava (Gradle):
implementation 'com.vibium:vibium:26.8.21'Java (Maven):
<dependency>
<groupId>com.vibium</groupId>
<artifactId>vibium</artifactId>
<version>26.8.21</version>
</dependency>This installs the Vibium binary and downloads Chrome automatically. No manual browser setup required.
Async API:
import { browser } from 'vibium'
import fs from 'node:fs/promises'
const bro = await browser.start()
const vibe = await bro.page()
await vibe.go('https://example.com')
const png = await vibe.screenshot()
await fs.writeFile('screenshot.png', png)
const link = await vibe.find('a')
await link.click()
await bro.stop()Sync API:
const { browser } = require('vibium/sync')
const fs = require('fs')
const bro = browser.start()
const vibe = bro.page()
vibe.go('https://example.com')
const png = vibe.screenshot()
fs.writeFileSync('screenshot.png', png)
const link = vibe.find('a')
link.click()
bro.stop()# Async
from vibium.async_api import browser
# Sync (default)
from vibium import browserAsync API:
import asyncio
from vibium.async_api import browser
async def main():
bro = await browser.start()
vibe = await bro.page()
await vibe.go("https://example.com")
png = await vibe.screenshot()
with open("screenshot.png", "wb") as f:
f.write(png)
link = await vibe.find("a")
await link.click()
await bro.stop()
asyncio.run(main())Sync API:
from vibium import browser
bro = browser.start()
vibe = bro.page()
vibe.go("https://example.com")
png = vibe.screenshot()
with open("screenshot.png", "wb") as f:
f.write(png)
link = vibe.find("a")
link.click()
bro.stop()var bro = Vibium.start();
var vibe = bro.page();
vibe.go("https://example.com");
var png = vibe.screenshot();
Files.write(Path.of("screenshot.png"), png);
var link = vibe.find("a");
link.click();
bro.stop();┌──────────────────────────────────────┐
│ LLM / Agent │
│ (Claude Code, Codex, Gemini, etc.) │
└──────────────────────────────────────┘
▲ ▲
│ CLI (Bash) │ MCP (stdio)
▼ ▼
┌───────────────────────────────────┐
│ Vibium binary │
│ │
│ ┌──────────────┐ ┌────────────┐ │
│ │ CLI Commands │ │ MCP Server │ │
│ └─────┬────────┘ └──────┬─────┘ │ ┌──────────────────┐
│ └───────▲─────────┘ │ │ │
│ │ │ │ │
│ ┌──────▼───────┐ │ BiDi │ Web Browser │
│ │ BiDi Proxy │ │◄──────►│ │
│ └──────────────┘ │ │ │
└───────────────────────────────────┘ └──────────────────┘
▲
│ WebSocket BiDi :9515
▼
┌──────────────────────────────────────┐
│ Client Libraries │
│ (js/ts | python | java) │
│ │
│ ┌─────────────────┐ ┌────────────┐ │
│ │ Async API │ │ Sync API │ │
│ │ await vibe.go() │ │ vibe.go() │ │
│ └─────────────────┘ └────────────┘ │
└──────────────────────────────────────┘
| Platform | Architecture | Status |
|---|---|---|
| Linux | x64 | ✅ Supported |
| macOS | x64 (Intel) | ✅ Supported |
| macOS | arm64 (Apple Silicon) | ✅ Supported |
| Windows | x64 | ✅ Supported |
Chrome is the default browser; Firefox is also supported (how to switch). Firefox 154+ can record video of a session.
See CONTRIBUTING.md for development setup and guidelines.
Source builds require Go, Node.js/npm, and Java 21 because the default make
target builds the Go binary, JS client, and Java client.
See ROADMAP.md for current development work and longer-term ideas.
Apache 2.0