PlugWiki

OpenCLI Browser Skill: Let Claude Drive Your Logged-In Chrome

opencli-browser lets Claude drive your own Chrome through the opencli command: read pages, fill forms, click and capture API calls on logged-in sites.

Installs
20.4K
on skills.sh
Stars
30.8K
on GitHub
Updated
Oct 2026
last repo commit

The skill teaches Claude the opencli browser commands: open a page, read its interactive elements, click, fill, type, select, upload, drag, wait, take screenshots and read page text. Work happens in named sessions (--session research), and results come back as one JSON value, so the agent can read success and error fields.

It can also list the network requests a page makes, with --static to add script-based candidates, and manage native Chrome tabs, groups and windows. bind attaches a tab you already have open. Its allowed tools are Bash(opencli:*), Read, Edit and Write.

It needs the OpenCLI command, the Browser Bridge Chrome extension and a small local daemon that starts when needed. Install those first, as the README describes, then add the skill.

Good for

  • People who want an agent to work inside sites where they are already logged in
  • Developers who need to capture a page's network calls to find its API
  • Anyone who wants Claude to manage Chrome tabs, groups and windows

Install opencli-browser

Claude Code

  1. 1

    Copy this prompt.

    Install the opencli-browser skill in Claude Code. Run this command:

npm install -g @jackwener/opencli

Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js 20.18.1 or later.

This skill drives the opencli command, so it needs OpenCLI and its Chrome extension first. On macOS and Windows the README recommends the OpenCLIApp desktop app (opencli.info/download) as an alternative to the npm install in step 1.

From the project README. Source.

What's in the skill

When it loads

Drive Chrome with OpenCLI to inspect pages, interact with logged-in sites, capture API evidence, and manage native tabs, groups and windows. For reusable adapters, see opencli-adapter-author.

Sections in SKILL.md

  1. Scope and results
  2. Page targets and interaction
  3. Network and scripts
  4. Native Chrome management

Key facts

Author
jackwener (GitHub: jackwener)
Repo
jackwener/opencli
Installs on skills.sh
20,382 as of 2026-11-02
GitHub stars (whole repo)
30,816 as of 2026-11-02
License
Apache-2.0
Last repo commit
2026-10-26
Files in the skill folder
1 file including SKILL.md
skills.sh audits
Gen Agent Trust Hub pass, Socket warn, Snyk warn

What it includes

  • --session <name>, --profile <name>

    Page commands need a session name. A profile picks a Chrome profile when you have several.

    Runs code · Sessions and profiles

  • List URL and interactive elements, locate by role, name or text

    state --diagnostics also reports the page title and initial-state globals. --source ax uses the accessibility tree.

    Runs code · state and find

  • click, fill, type, select, check, upload, drag, keys, dialog

    Acts on a page in your Chrome. fill replaces a field and verifies the value.

    Runs code · Interaction commands

  • get, extract, screenshot, eval

    eval runs JavaScript in the page and can change state, per SKILL.md.

    Runs code · Read commands

  • Wait for xhr, selector, text or download; manage session tabs

    tab list, new, select and close. frames lists frame targets.

    Runs code · wait and tab

  • Attach or release a tab you already have open

    Bound sessions allow navigation but block creating, selecting or closing tabs.

    Runs code · bind and unbind

  • List requests, details, filters, raw bodies, static script analysis

    --raw exposes response bodies. --static adds candidates found in loaded scripts, which SKILL.md says are not executed or proven APIs.

    Runs code · network

  • Native Chrome tabs, tab groups and windows

    Calls such as tabs.query, tabs.group and tabGroups.update, with no page attached.

    Runs code · chrome describe / call

From the official docs

Permissions

High risk

This skill lets Claude act inside your real, logged-in Chrome, so it can click, type and submit on your behalf. Use a separate Chrome profile and review writes before approving them.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It acts as you on logged-in sites

    Click, fill, type, upload and eval run in your real browser session. A wrong action can post, buy or delete. Use a dedicated Chrome profile and confirm actions that change data. github.com

  • eval runs JavaScript in the page

    SKILL.md says eval can mutate state. Approve each use and read the code first. github.com

  • Network capture can expose response bodies

    network --raw shows raw response bodies, which can include tokens and personal data from pages you are logged in to. github.com

  • A local daemon and a browser extension sit between Claude and Chrome

    The README says OpenCLI connects to Chrome through the Browser Bridge extension plus a small local daemon that auto-starts when needed. Anything that can reach the daemon can act through your logged-in browser. github.com

  • Setup installs a global package and an extension

    The README has you run npm install -g @jackwener/opencli and add a Chrome extension, possibly in Developer mode. Install the extension from the Chrome Web Store when you can. github.com

  • Native Chrome calls reach tabs, groups and windows

    chrome call can run tabs.query across your Chrome (the SKILL.md example uses '[{}]') and change tab groups with tabs.group and tabGroups.update. SKILL.md says mutations touching OpenCLI-managed tabs return managed_resource_conflict, and tabs or windows created natively persist. github.com

Technical details
Talks tosites you open in Chrome
Scopes

  • your logged-in Chrome profile

  • OpenCLI Browser Bridge extension

  • local OpenCLI daemon

Questions

What does the opencli-browser skill do?

It teaches Claude to operate a real Chrome through the opencli browser commands: inspect pages, interact with sites you are logged in to, capture network calls and manage tabs, groups and windows.

How do I install it?

Use the install box above. It lists the whole setup from the README: the OpenCLI command, the Chrome extension, a check with opencli doctor, then the skill. The skill alone does nothing without the command.

Does it work in the Claude app?

The README says to install the skill in an agent such as Claude Code or Cursor, using the skills CLI. The skill runs the opencli command, so it needs an agent that can run commands on your machine.

Does it need an API key?

SKILL.md names none. It works through your own logged-in Chrome session, using the OpenCLI Browser Bridge extension and a small local daemon that the README says auto-starts.

Is it safe to use?

It acts as you on every site where you are logged in, so a wrong click or form fill can post, buy or delete. SKILL.md says to inspect the affected state before retrying a write when an error outcome is unknown. Our advice: use a separate Chrome profile for agent work and watch each write.

How is it different from opencli-adapter-author?

This skill drives pages ad hoc. The description points to opencli-adapter-author for reusable adapters, and the README says that skill writes a new adapter end to end.

Information

updated Nov 2, 2026

Repository
jackwener/opencli
Folder
skills/opencli-browser

jackwener/opencliskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2
  1. OpenCLI README · github.com
  2. opencli-browser SKILL.md · github.com

Updated Nov 2, 2026.