Gbrow — The Browser Your AI Agent Actually Needs#

A full-featured headless browser powered by Playwright and Bun. Uses the accessibility tree for page reading — not expensive vision models.

Why Gbrow?#

Traditional (screenshots + vision)Gbrow (accessibility tree)Screenshot → upload to GPT-4o → wait → readariaSnapshot() → instant structured text~$0.01 per page readFree3-10 seconds per page**< 100msFails on API key issuesAlways worksClick by fragile CSS selectorClick by @ref** ( @e1, @e2, etc.)

Quick Setup#

bash

# Clone and install
git clone https://github.com/ashish797/Gbrow.git ~/.openclaw/workspace/skills/Gbrow
cd ~/.openclaw/workspace/skills/Gbrow
bash setup.sh

Or one-liner:

bash

curl -fsSL https://raw.githubusercontent.com/ashish797/Gbrow/main/setup.sh | bash

How It Works#

1. Start the server#

bash

cd ~/.openclaw/workspace/skills/Gbrow
bun run src/server.ts

2. Read the page (accessibility tree)#

The snapshot gives you a structured view with clickable refs:

text

@e1 [heading] "Welcome" [level=1]
@e2 [link] "Get Started"
@e3 [button] "Sign in"
@e4 [textbox] "Search"

3. Click by ref#

text

click @e2     → clicks "Get Started"
fill @e4 "query"  → types into search box

Commands#

Navigation#

CommandDescriptionExamplegoto <url>Navigate to URLgoto https://example.com``backHistory backback``forwardHistory forwardforward``reloadReload pagereload``urlPrint current URLurl

Reading#

CommandDescriptionExamplesnapshotAccessibility tree with @refssnapshot -i (interactive only)textCleaned page texttext``html [selector]Raw HTMLhtml .article``linksAll links as "text → href"links``formsForm fields as JSONforms

Interaction#

CommandDescriptionExampleclick <ref>Click elementclick @e3``fill <ref> <text>Fill inputfill @e4 "hello"``select <ref> <value>Select dropdownselect @e5 "option1"``type <ref> <text>Type with keyboardtype @e4 "search term"``press <key>Press keypress Enter``scroll <direction>Scroll pagescroll down

Inspection#

CommandDescriptionExamplejs <expr>Run JavaScriptjs document.title``css <sel> <prop>Computed CSScss .box color``attrs <ref>Element attributesattrs @e1``is <prop> <ref>State checkis visible @e3

Tabs#

CommandDescriptiontabsList open tabstab NSwitch to tab NnewtabOpen new tabclosetabClose current tab

Visual#

CommandDescriptionscreenshotTake screenshotresponsive <w> <h>Set viewport sizepdfSave page as PDF

Snapshot Flags#

FlagDescription-iInteractive elements only (buttons, links, inputs)-cCompact (remove empty structural nodes)-d NLimit tree depth-s <sel>Scope to CSS selector-DDiff against previous snapshot-aAnnotated screenshot with ref overlays

HTTP API#

All commands go through the HTTP API:

bash

# Get port and token from state file
PORT=$(python3 -c "import json; print(json.load(open('.gstack/browse.json'))['port'])")
TOKEN=$(python3 -c "import json; print(json.load(open('.gstack/browse.json'))['token'])")

# Send command
curl -s -X POST "http://127.0.0.1:${PORT}/command" \
  -H "Authorization: Bearer ${TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"command":"goto","args":["https://example.com"]}'

Architecture#

text

┌─────────────┐     HTTP      ┌──────────────────┐
│  OpenClaw   │ ──────────▶  │  Gbrow Server    │
│  Agent      │              │  (Bun + Playwright)│
└─────────────┘              └────────┬─────────┘
                                      │
                                      ▼
                              ┌──────────────────┐
                              │  Chromium         │
                              │  (headless)       │
                              └──────────────────┘
                                      │
                                      ▼
                              ┌──────────────────┐
                              │ Accessibility     │
                              │ Tree (ariaSnapshot)│
                              └──────────────────┘

No vision models. No API calls. Just structured text from the browser's accessibility layer.

Credits#

Built on top of gstack by Gary Tan (Y Combinator). Adapted for OpenClaw with permission under MIT license.

License#

MIT