feat: Added in-built browser tools using playwright and a parallel playwright node process using spawn for bun-playwright issues.

[forloopcodes]

Added browser automation toolkit to the packages/opencode package, just like anthropic's claude in chrome and cursor 2.0. Also, this feature can be turned on by setting the flag OPENCODE_ENABLE_BROWSER=true in the terminal or setting browser: true in opencode.json before running it.

This method does not use mcp but direct integration of browser tools into opencode, doing this can further extend opencode's possibilities to open devtools (not yet, but possible), resize windows, and much more than what extension+mcp method does.

Dependencies:

• playwright

Working video:

opencode.-.added.browser.tools.mp4

(working speed depends on network ping, unfortunately mine is too high)

Tools:

File	Tool Name	Description
navigate.ts	browser_navigate	Navigate to URLs with optional content return
navigate-back.ts	browser_navigate_back	Go back in browser history
navigate-back.ts	browser_navigate_forward	Go forward in browser history
wait.ts	browser_wait	Wait for conditions (time, text, elements)
init.ts	browser_init	Initialize browser instance with optional direct URL navigation
close.ts	browser_close	Close browser instance
close-page.ts	browser_close_page	Close current page while keeping browser running
click.ts	browser_click	Click elements with fuzzy search and auto-scroll
hover.ts	browser_hover	Hover over elements
type.ts	browser_type	Type text into inputs with submit option
search.ts	browser_search	Find elements using fuzzy search
drag.ts	browser_drag	Drag and drop between elements
scroll.ts	browser_scroll	Scroll page with content return
press-key.ts	browser_press_key	Press keyboard keys
fill-form.ts	browser_fill_form	Fill multiple form fields at once
select-option.ts	browser_select_option	Select dropdown options
check-element.ts	browser_check	Check/uncheck checkboxes and radio buttons
file-upload.ts	browser_file_upload	Upload files to file inputs
content.ts	browser_content	Get page content (text, links, inputs, structured)
screenshot.ts	browser_screenshot	Take screenshots (viewport or full page)
snapshot.ts	browser_snapshot	Get accessibility snapshot for actions
evaluate.ts	browser_evaluate	Evaluate JavaScript expressions
run-code.ts	browser_run_code	Run Playwright code snippets
console-messages.ts	browser_console_messages	Get browser console messages
network-requests.ts	browser_network_requests	List network requests
handle-dialog.ts	browser_handle_dialog	Handle browser dialogs (alerts, confirms)
tabs.ts	browser_tabs	Manage browser tabs (list, create, close, select)
resize.ts	browser_resize	Resize browser window
get-page.ts	browser_get_page	Get current page information (URL, title, scroll position)
get-element-at.ts	browser_get_element_at	Get element at specific coordinates (x, y)
get-element-bounds.ts	browser_get_element_bounds	Get bounding box of element by selector/ref
verify-element-visible.ts	browser_verify_element_visible	Verify elements are visible
verify-text-visible.ts	browser_verify_text_visible	Verify text appears on page
generate-locator.ts	browser_generate_locator	Generate robust selectors for tests

I've made sure these tools give expandable web automation capabilities and infinite possibilities with minimal tool call overhead and context bloat.

Fixes:

• [FEATURE]: Browser automation like Claude in Chrome #6992
• BrowserTools and Automation #575

Known issues:

• need to do npx playwright install chromium before getting started.

Notes:

• this is my first PR on the opencode repository, would love to listen from the community
• developers/reviewers are requested to tell me whatever i could add/fix over this feature (im sorry if something broke) <3

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

No duplicate PRs found

[Copilot]

Pull request overview

This PR adds a comprehensive browser automation toolkit to OpenCode using Playwright, providing 28 tools for web automation tasks. The implementation uses a Node.js HTTP server bridge (browser-server.js) to manage Chromium instances, as Playwright doesn't work natively with Bun. The feature is gated behind the OPENCODE_ENABLE_BROWSER environment flag.

Key Changes:

• Browser automation infrastructure: BrowserManager, HTTP bridge server, and 28 automation tools
• Dependencies: Added playwright@1.57.0 and sharp@0.34.5
• Configuration: TypeScript path aliases, new environment flags, and tool registry integration

Reviewed changes

Copilot reviewed 39 out of 41 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
tsconfig.json	Added esModuleInterop and path aliases for imports
packages/opencode/package.json	Added playwright and sharp dependencies
packages/opencode/src/flag/flag.ts	Added OPENCODE_ENABLE_BROWSER and OPENCODE_BROWSER_PROFILE_PATH flags
packages/opencode/src/tool/tool.ts	Added Tool.attachExecute helper for convenience execute method
packages/opencode/src/tool/registry.ts	Integrated BrowserTools into tool registry
packages/opencode/src/browser/manager.ts	Core BrowserManager with HTTP-based Playwright bridge
packages/opencode/src/browser/tools/*.ts	28 individual browser automation tools
packages/opencode/browser-server.js	Node.js HTTP server managing Playwright browser instance
packages/opencode/src/browser/annotate.ts	Screenshot annotation utilities using sharp
packages/opencode/src/browser/README.md	Comprehensive documentation
bun.lock	Lock file updates for new dependencies

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[gangjeuk]

How do you think about the Puppeteer and Puppeteer chromium resolver combination?

[forloopcodes]

i didnt think a lot about choosing between them, but one advantage is that playwright can control multi tabs in a single instance (cant confirm, source: gemini)

[github-actions]

Hey! Your PR title [FEATURE] Added in-built browser tools using playwright and a parallel playwright node process using spawn for bun-playwright issues. doesn't follow conventional commit format.

Please update it to start with one of:

• feat: or feat(scope): new feature
• fix: or fix(scope): bug fix
• docs: or docs(scope): documentation changes
• chore: or chore(scope): maintenance tasks
• refactor: or refactor(scope): code refactoring
• test: or test(scope): adding or updating tests

Where scope is the package name (e.g., app, desktop, opencode).

See CONTRIBUTING.md for details.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[h0jeZvgoxFepBQ2C]

Cant wait for this merge!

[BrutchsamaJeanLouis]

THIS!!! 🎉

[dj0024javia]

bumping this. really helpful! waiting for the merge 🚀

[transportrefer]

Any status on this PR? This would be really nice.

[SpartanJohn117]

Hi @forloopcodes,

Love this. What is the current status 🤔

[skywalker-35]

This is exactly what OpenCode needs. Having built-in browser tools (not MCP, not external CLI) as first-class citizens would be a huge win for the agent workflow.

Currently I'm hopping between agent-browser, playwright-cli, and Playwright MCP — each has friction (MCP token bloat, CLI cold starts, 360 antivirus blocking Chrome on Windows). Native integration via browser: true in opencode.json is the cleanest path forward.

Looking forward to seeing this merged. Happy to help test on Windows if needed.

[kacperpaczos]

I'm waiting for this feature to be available.