jb-chrome-mcp

name: jb-chrome-mcp description: Use when the user asks for Chrome MCP/DevTools MCP via mcporter, existing Chrome tabs, console/network inspection, screenshots, or performance traces. homepage: https://github.com/ChromeDevTools/chrome-devtools-mcp metadata: {"clawdbot":{"requires":{"bins":["mcporter"]},"install":[{"id":"node","kind":"node","package":"mcporter","bins":["mcporter"],"label":"Install mcporter (node)"}]}} skill_author: bjesuiter@gmail.com private: true

JB Chrome MCP

Private skill for using Chrome DevTools MCP through mcporter against the user's existing Chrome session.

This skill defines JB-specific rules, bootstrap, safety, and recovery workflow only. For the full upstream tool catalog and exact arguments, read references/chrome-devtools-mcp-tool-reference.md instead of duplicating it here.

Hard rule: reattach to the existing Chrome profile only. Do not silently switch to isolated Chrome, Playwright, Puppeteer, the Codex in-app browser, AppleScript, osascript, GUI scripting, or macOS open for browser control unless the user explicitly asks for an isolated/new browser.

Screenshot/live UI bugs require this existing-Chrome path. curl, source inspection, Worker smoke tests, or local Playwright are supporting proof only; do not treat them as equivalent when the user showed a rendered browser problem or the page may depend on login/profile state.

Prerequisites

Before using this skill:

1. Open chrome://inspect/#remote-debugging in Chrome.
2. Turn remote debugging on and allow incoming debugging connections in Chrome's dialog.
3. Confirm Chrome is listening on 127.0.0.1:9222.
   • Expected listener: Server running at: 127.0.0.1:9222
   • Useful check: lsof -nP -iTCP:9222 -sTCP:LISTEN
4. Make sure mcporter is installed.

Bootstrap rule

Do not ask the user to configure mcporter manually unless the bootstrap script fails.

Before the first mcporter call, run the bundled script. From this skills repo:

skills/jb-chrome-mcp/scripts/ensure-jb-chrome-mcp.sh

From an installed skill path:

/Users/bjesuiter/.agents/skills/jb-chrome-mcp/scripts/ensure-jb-chrome-mcp.sh

The script:

• verifies Chrome is listening on 127.0.0.1:9222
• installs or updates a saved home-level mcporter server named jb-chrome-mcp
• configures chrome-devtools-mcp@1.2.0 with --auto-connect
• passes --userDataDir "$HOME/Library/Application Support/Google/Chrome" so DevTools MCP reattaches to the user's real Chrome profile and tabs

--auto-connect plus the real Chrome --userDataDir is more reliable for the chrome://inspect/#remote-debugging workflow than forcing --browserUrl http://127.0.0.1:9222.

After bootstrapping, verify the tool list if needed:

mcporter list jb-chrome-mcp --schema

This creates a saved home-level mcporter server named jb-chrome-mcp. That is what lets later calls use short names like mcporter call jb-chrome-mcp.list_pages instead of repeating the full npx chrome-devtools-mcp ... stdio config every time.

Expected saved transport:

npx -y chrome-devtools-mcp@1.2.0 --auto-connect --userDataDir "$HOME/Library/Application Support/Google/Chrome" --no-usage-statistics

Note: the first live mcporter call may trigger a Chrome permission prompt for the DevTools MCP connection. If a live call hangs, check Chrome for an Allow prompt before changing config or restarting processes.

Preferred workflow

1. list_pages
2. Confirm the list shows the user's real open tabs. If it shows a blank/default isolated Chrome, stop and say reattach failed.
3. select_page
4. take_snapshot before interacting. Use current uid values only from the latest snapshot.
5. Pick the tool from references/chrome-devtools-mcp-tool-reference.md:
   • Prefer fill_form for multi-field forms.
   • Use handle_dialog for page-level alert/confirm/prompt; use attach prompt recovery below only for native Chrome permission prompts.
   • For performance, use trace tools for performance metrics; Lighthouse excludes performance.
   • Do not use experimental-only tools (click_at, screencast) unless the server was configured with their required flags.

Good defaults

• Prefer mcporter call ... --args '{...}' --output json for machine-readable responses.
• Prefer take_snapshot over screenshots when text structure is enough.
• Call list_pages first if the active tab is unclear.
• Use select_page with bringToFront: true before interactive actions when tab focus matters.
• For first live calls, use mcporter's built-in timeout so the agent gets a structured failure instead of hanging forever on a Chrome permission prompt:

mcporter call jb-chrome-mcp.list_pages --args '{}' --output json --timeout 20000

Do not wrap mcporter in shell timeout for normal use; it can kill the client without a useful MCP error.

• Avoid noisy recovery loops. Repeated MCP/browser restarts can trigger reconnect/login prompts and alerts. Try one recovery path, then pause and ask or report the verification gap.

Command pattern

First run the bootstrap script, then use the saved server:

./scripts/ensure-jb-chrome-mcp.sh
mcporter call jb-chrome-mcp.<tool_name> --args '<json-args>' --output json --timeout 20000

Use references/chrome-devtools-mcp-tool-reference.md for tool names, parameters, examples, and flag requirements.

Fallback for explicit browser URL mode

If --autoConnect is unavailable or you are connecting to an older/manual Chrome instance that exposes /json/version, use:

mcporter config add jb-chrome-mcp \
  --stdio npx \
  --arg -y \
  --arg chrome-devtools-mcp@1.2.0 \
  --arg --browserUrl \
  --arg http://127.0.0.1:9222 \
  --arg --userDataDir \
  --arg "$HOME/Library/Application Support/Google/Chrome" \
  --arg --no-usage-statistics \
  --scope home

Attach prompt recovery

If list_pages hangs while Chrome shows an auth/attach/update prompt, handle the attach alert before falling back. Prefer Peekaboo to press an explicit Chrome Allow button when visible; otherwise wait for the human. Do not restart daemons or kill MCP processes just because the first output is slow.

PB="${PEEKABOO_BIN:-$HOME/bin/peekaboo}"
[ -x "$PB" ] || PB="$(command -v peekaboo)"
"$PB" permissions status --json
"$PB" see --app frontmost --path /tmp/chrome-attach.png --json --annotate
# If the UI shows Chrome "Allow remote debugging?", click only the visible Allow button.
"$PB" click --coords <allow_x>,<allow_y> --json
mcporter call jb-chrome-mcp.list_pages --args '{}' --output json --timeout 20000

Use coordinates from the current Peekaboo snapshot, not stale notes. Success means list_pages returns the user's real Chrome tabs.

Attach-alert rule: when the current snapshot clearly shows Chrome asking to allow DevTools/MCP/browser automation attachment, click the visible Allow button once, then rerun list_pages. If the button is not visible or the prompt is ambiguous, stop and ask; do not silently switch to Playwright/Puppeteer.

Secret handling

Never print tokens/passwords from page DOM, network logs, or inputs. For token checks, return shape only: present/absent, length, status code, account/org name.

get_network_request can return request/response bodies inline. Prefer file output args from the reference for large or potentially sensitive bodies, then summarize only safe shape/status details.

Avoid debug logging by default. Do not enable DEBUG=*, --logFile, or verbose protocol logging unless explicitly debugging MCP internals and prepared to redact the output. Chrome DevTools MCP debug logs can include live URLs, request headers, cookies, auth tokens, post bodies, and other sensitive browser-session data from open tabs.

Troubleshooting

• If Chrome is not listening on 9222, reopen chrome://inspect/#remote-debugging and turn it on again.
• If list_pages hangs, check Chrome for an Allow prompt and use attach prompt recovery above.
• If list_pages fails with DevToolsActivePort, ask the user to restart Chrome or the DevTools bridge, then retry once:

mcporter daemon restart
mcporter call jb-chrome-mcp.list_pages --args '{}' --output json --timeout 20000

• If list_pages returns a connection error, retry with the persisted config form instead of ad-hoc stdio.
• If --browserUrl fails on http://127.0.0.1:9222/json/version, switch back to --autoConnect.
• If the wrong tab is active, run list_pages and select_page again.
• If browser automation is unavailable, report that as a verification gap instead of substituting isolated browser tooling.