chrome-devtools-mcpSAFE
Chrome DevTools for coding agents
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://npmjs.org/package/chrome-devtools-mcp)
Chrome DevTools for agents (chrome-devtools-mcp) lets your coding agent (such as Antigravity, Claude, Cursor or Copilot) control and inspect a live Chrome browser. It acts as a Model-Context-Protocol (MCP) server, giving your AI coding assistant access to the full power of Chrome DevTools for reliable automation, in-depth debugging, and performance analysis. A [CLI][cli] is also provided for use without MCP.
[Tool reference][tool-reference] | [Changelog][changelog] | [Contributing][contributing] | [Troubleshooting][troubleshooting] | [Design Principles][design-principles]
Key features
- Get performance insights: Uses [Chrome
DevTools](https://github.com/ChromeDevTools/devtools-frontend) to record traces and extract actionable performance insights.
- Advanced browser debugging: Analyze network requests, take screenshots and
check browser console messages (with source-mapped stack traces).
- Reliable automation. Uses
puppeteer to automate actions in Chrome and automatically wait for action results.
Disclaimers
chrome-devtools-mcp exposes content of the browser instance to the MCP clients allowing them to inspect, debug, and modify any data in the browser or DevTools. Avoid sharing sensitive or personal information that you don't want to share with MCP clients.
chrome-devtools-mcp officially supports Google Chrome and Chrome for Testing only. Other Chromium-based browsers may work, but this is not guaranteed, and you may encounter unexpected behavior. Use at your own discretion. We are committed to providing fixes and support for the latest version of Extended Stable Chrome.
Performance tools may send trace URLs to the Go
122de45ea344OBSERVED · 2026-10-20Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add chrome-devtools-mcp -- npx -y [email protected]
Exposed tools (93)
79 read · 12 write · 2 destructive. Blast radius: 2 tools can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
3pDeveloperTool | read | A test tool |
browserless_tool | read | A tool that does not need a browser |
click | read | Clicks on the provided element |
click_at | read | Clicks at the provided coordinates |
close_heapsnapshot | read | Closes a previously loaded and cached memory heapsnapshot, freeing its memory. |
close_page | read | Closes the page by its index. The last open page cannot be closed. |
compare_heapsnapshots | read | Loads and caches two memory heapsnapshots and returns the comparison. If classIndex is provided, returns detailed diff for that class, otherwise returns summary diff. |
custom_tool | read | A custom tool |
debug_tool | read | A debugging WebMCP tool |
disabled_tool | read | A disabled tool |
drag | read | Drag an element onto another element |
dynamic_debug_tool | read | Added dynamically |
emulate | read | Emulates various features on the target page. Each call replaces all emulation settings (except extraHttpHeaders), so pass all desired settings in a single call. |
enum_tool | read | test description |
evaluate | read | Evaluates a JavaScript script |
evaluate_script | read | Evaluate JavaScript inside the target page${cliArgs.categoryExtensions ? |
execute_3p_developer_tool | write | Executes a third-party developer tool exposed by the page. |
execute_webmcp_tool | write | Executes a WebMCP tool exposed by the page (for debugging tools, use |
file_tool | read | A tool requiring file validation |
fill | read | Type text into an input, text area or select an option from a <select> element. |
fill_form | read | Fill out multiple form elements (inputs, selects, checkboxes, radios) at once. ALWAYS prefer this tool over multiple individual |
get_console_message | read | Gets a console message by its ID. You can get all messages by calling ${LIST_CONSOLE_MESSAGES_TOOL_NAME}. |
get_devtools_comments | read | Retrieve user comments from the DevTools window for the page. |
get_heapsnapshot_class_nodes | read | Loads and caches a memory heapsnapshot and returns instances of a specific class with their IDs. |
get_heapsnapshot_details | read | Loads and caches a memory heapsnapshot and returns all available information including statistics, static data, and aggregated node information. Supports pagination for aggregates. |
get_heapsnapshot_dominators | read | Loads and caches a memory heapsnapshot and returns the dominator chain for a specific node ID. This helps to identify which objects are keeping the target node alive. |
get_heapsnapshot_duplicate_strings | read | Loads and caches a memory heapsnapshot and returns duplicate strings grouped by their value. |
get_heapsnapshot_edges | read | Loads and caches a memory heapsnapshot and returns outgoing edges (references) for a specific node ID. |
get_heapsnapshot_object_details | read | Loads and caches a memory heapsnapshot and returns detailed information about a specific object by its node ID, including size, type, distance, and DOM detachedness. |
get_heapsnapshot_retainers | read | Loads and caches a memory heapsnapshot and returns retainers for a specific node ID. |
get_heapsnapshot_retaining_paths | read | Loads and caches a memory heapsnapshot and returns retaining paths for a specific node ID. This helps to understand why a node is not being garbage collected. |
get_heapsnapshot_summary | read | Loads and caches a memory heapsnapshot and returns snapshot summary stats, including native contexts and their sizes, and retained by context summary. |
get_network_request | read | Gets a network request by an optional reqid, if omitted returns the currently selected request in the DevTools Network panel. Useful for inspecting request headers (including |
get_os_app_state | read | Returns the OS integration state (badge count and registered file |
get_tab_id | read | Get the tab ID of the page |
global_tool | read | A global tool |
group-1 | read | desc-1 |
group-2 | read | desc-2 |
handle_dialog | read | If a browser dialog was opened, use this command to handle it |
hanging_context_tool | read | A tool whose browser connection never resolves |
hover | read | Hover over the provided element |
hybrid_tool | read | A tool with both schema file verifications |
install_extension | write | Installs a Chrome extension from the given path. |
install_pwa | write | Installs a Progressive Web App (PWA) identified by its manifest ID. |
install_pwa_tool | write | PWA tool with local-only file verification |
launch_pwa | read | Launches an installed Progressive Web App using its saved display mode. |
legacy_webmcp_tool | read | A WebMCP tool without debugging annotation |
lighthouse_audit | read | Get Lighthouse scores, failing audits, and the DOM nodes they flag for accessibility, SEO, best practices, and agentic browsing. Full reports are saved to disk. This excludes performance. For performance audits, run ${startTrace(args).name} |
list_3p_developer_tools | read | Lists all third-party developer tools the page exposes for providing runtime information (including debugging WebMCP tools and tools exposed via the |
list_extensions | read | Lists all the Chrome extensions installed in the browser. This includes their name, ID, version, and enabled status. |
list_network_requests | read | Lists the most recent requests for the target page since the last navigation. |
list_pages | read | Get a list of pages${args?.categoryExtensions ? |
list_webmcp_tools | read | Lists all WebMCP tools the page exposes (excluding debugging tools, which are exposed via |
myTool | read | Does something |
navigate | read | Loads a URL |
navigate_page | read | Go to a URL, or back, forward, or reload. Use project URL if not specified otherwise. |
new_page | read | Open a new tab and load a URL. Use project URL if not specified otherwise. |
open_devtools | read | Open a DevTools window for the selected page. |
page_file_tool | read | A page scoped tool with file verification |
page_tool | read | A page scoped tool |
performance_analyze_insight | read | Provides more detailed information on a specific Performance Insight of an insight set that was highlighted in the results of a trace recording. |
performance_start_trace | write | Start a performance trace on the target webpage. Use to find frontend performance issues, Core Web Vitals (LCP, INP, CLS), and improve page load speed. |
performance_stop_trace | write | Stop the active performance trace recording on the target webpage. |
press_key | read | Press a key or key combination. Use this when other input methods like fill() cannot be used (e.g., keyboard shortcuts, navigation keys, or special key combinations). |
query_heapsnapshot_objects | read | Loads and caches a memory heapsnapshot and queries objects matching specific filters (className, propertyName, nodeType, retainedSize, selfSize, isDetached, sortBy). |
reload_extension | read | Reloads an unpacked Chrome extension by its ID. |
remote_file_tool | read | A tool with remote-only file verification |
resize_page | read | Resizes the page |
resolve_devtools_comment | read | Append an agent reply to a DevTools comment thread and mark it as resolved. |
reveal_in_devtools | read | Navigate DevTools to a specified panel and highlight a target DOM node or network request. The parameters uid and reqid are mutually exclusive. |
screencast_start | write | Starts recording a screencast (video) of the target page. |
screencast_stop | write | Stops the active screencast recording on the target page. |
screenshot | read | Takes a screenshot |
select_page | read | Select a page as a context for future tool calls. |
slow_tool | read | A tool whose handler legitimately outlasts the timeout |
strict_tool | read | A tool with a required argument |
take_heapsnapshot | read | Capture a heap snapshot of the target page. Use to analyze the memory distribution of JavaScript objects and debug memory leaks. |
take_screenshot | read | Take a screenshot of the page or element. |
test-group | read | test description |
test-tool | read | test tool description |
test_tool | read | A test tool |
tool-1 | read | tool-1-desc |
tool-2 | read | tool-2-desc |
trigger_extension_action | write | Triggers the default action of an extension by its ID. |
type_text | read | Type text using keyboard into a previously focused input |
uninstall_extension | destructive | Uninstalls a Chrome extension by its ID. |
uninstall_pwa | destructive | Uninstalls a Progressive Web App identified by its manifest ID and |
upload_file | write | Upload a file through a provided element. |
upload_tool | write | A tool with local-only file verification |
user_tool | read | A non-debugging WebMCP tool |
wait_for | read | Wait for the specified text to appear on the selected page. |
webmcp_tool | read | A WebMCP tool |
wrapped_enum_tool | read | test description |
Trust audit
SAFEgrade B · trust 84/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (9 observation(s))
- Network
- declared (4 observation(s))
- Shell
- declared (5 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (25)
uninstall_extension, uninstall_pwa
basic-trace.json.gz
.gitallowed
.gitmodules
.prettierignore
.release-please-manifest.json
let isChecked = await page.$eval(
isChecked = await page.$eval(
let swChecked = await page.$eval(
swChecked = await page.$eval(
let r1Checked = await page.$eval(
import {logger} from '../../utils/logger.js';import {logger, flushLogs, saveLogsToFile} from '../../utils/logger.js';import '../../third_party/devtools-frontend/front_end/entrypoints/formatter_worker/formatter_worker-entrypoint.js';
import '../../third_party/devtools-frontend/front_end/entrypoints/heap_snapshot_worker/heap_snapshot_worker-entrypoint.js';
export * as DevTools from '../../third_party/devtools-frontend/mcp/mcp.js';
Add the `--browser-url` option to your MCP client configuration. The value of this option should be the URL of the running Chrome instance. `http://127.0.0.1:9222` is a common default.
"--browser-url=http://127.0.0.1:9222"
"--browser-url=http://127.0.0.1:9222"
4. Select **Standard IO** and set **Run command** to `npx`. Use **Add argument** to enter `-y`, `chrome-devtools-mcp@latest`, and `--browser-url=http://127.0.0.1:9222` as separate arguments, adjusting
- **Connection URL:** `http://127.0.0.1:8080/mcp`
@blackwell-systems/gcf, @google/genai, @rollup/plugin-commonjs, @rollup/plugin-json, @rollup/plugin-node-resolve, @stylistic/eslint-plugin, @toon-format/toon, @types/filesystem
src/third_party/lighthouse-devtools-mcp-bundle.js
tests/fixtures/example.heapsnapshot
tests/fixtures/heap-1.heapsnapshot
Gates applied: no_behavioural_pass.
122de45ea344full audit observations/trust-audit/mcp-server/chromedevtools__chrome-devtools-mcp.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-20 | 122de45ea344 | SAFE | B | 84 | first audit |
Questions
What is the chrome-devtools-mcp MCP server?
Chrome DevTools for coding agents
What tools does chrome-devtools-mcp expose?
93 in total: 79 read-only, 12 that write, and 2 that can delete or overwrite (uninstall_extension, uninstall_pwa). Every one is listed on this page with its risk.
Is chrome-devtools-mcp safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (84/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B. Separately from the audit: 2 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does chrome-devtools-mcp need?
It reads GEMINI_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does chrome-devtools-mcp run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as chrome-devtools-mcp at 1.11.0.
How current is this page?
The grade is for one exact copy of the source (122de45ea344), read on 2026-10-20. The repository is watched and re-audited when it changes.
Add this badge to your README
Shows this page's audit result and links back to it. It updates when the audit does.
[](https://modelsatlas.com/mcp/chromedevtools/chrome-devtools-mcp/)
Not your listing, or something out of date? Claim or update your listing.