To use a browser with Hermes Agent, connect a browser-capable MCP server or attach Hermes to a local Chromium-family browser through CDP. For a Windows Chrome session controlled from WSL2, the Hermes guide recommends the chrome-devtools-mcp bridge. MCP adds tools to Hermes; it does not replace Hermes as the agent. Start with one browser route, confirm which tools Hermes can see, and limit access to only the actions you need.
How browser MCP fits into Hermes
Model Context Protocol (MCP) is an adapter layer: Hermes remains the agent that interprets your request, while an MCP server supplies tools the agent can call. Hermes discovers those tools when it starts or reloads MCP configuration. A browser MCP server can expose browser actions; the exact available actions depend on the server and backend you configure.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
Hermes describes browser pages as accessibility-tree snapshots with reference IDs for interactive elements. That means the agent works from a structured view of page content and controls rather than relying only on a screenshot. MCP support is included in a standard Hermes installation, and Hermes supports local servers that communicate over stdio as well as remote HTTP MCP servers.
Choose a browser route
Choose based on where the browser should run, whether you need an existing signed-in session, and whether the target is private or needs managed anti-bot capabilities. Do not connect several backends just to get started; one working route makes it easier to understand which tools and session the agent is using.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
| Route | Use it when | Important distinction |
|---|---|---|
| Browser Use with local packaged Chromium | You want Hermes to drive a local browser and do not need to attach to a browser profile already in use. | Browser Use mode is the default when configured. The packaged Chromium can be installed through Hermes tools. |
| Browser Use with a cloud backend | You need managed browser sessions, or a cloud route better fits the target site. | Browser Use cloud provides managed Chromium, stealth, residential proxies, CAPTCHA solving, and persistent profiles. Browserbase and Firecrawl are alternative cloud providers. |
| Local CDP connection | You want to use a Chromium-family browser in the same environment, especially one where you are already signed in. | /browser connect attaches through Chrome DevTools Protocol (CDP); it is an interactive Hermes CLI command. |
| Windows Chrome from WSL2 | Hermes runs in WSL2 while your signed-in Chrome runs on Windows. | The Hermes guide recommends the chrome-devtools-mcp stdio bridge for this cross-environment setup. |
When choosing among these routes, weigh local versus cloud execution, whether existing cookies matter, anti-bot needs, network access to private URLs, session isolation, and operating cost. The documentation does not provide a comparative price or performance table for the named cloud providers, so those should be checked with each provider rather than inferred from the feature descriptions.
Connect a local browser through CDP
For a local Chrome, Brave, Chromium, or Edge browser, Hermes can connect to the default CDP address http://127.0.0.1:9222 or to a specific WebSocket endpoint. This is the direct route when Hermes and the browser are in the same environment and you want the browser session already open to the user.
- Open Hermes CLI. The connect, status, and disconnect commands are interactive CLI commands. They are not dispatched by gateway chats such as WebUI, Telegram, or Discord.
- Connect. Enter
/browser connectto let Hermes tryhttp://127.0.0.1:9222. If your browser exposes CDP at another endpoint, provide its specificws://host:portendpoint with the command. - Check the session. Run
/browser statusand confirm Hermes reports the connection. - Use the browser tools. Ask Hermes to inspect a page or interact with an element, then verify that it is using the expected browser session.
- Detach when finished. Run
/browser disconnectto detach Hermes from the browser.
Hermes may auto-launch a supported browser with remote debugging. If you launch a browser manually, the Hermes guidance is to use a dedicated user-data directory. That keeps the debugging session distinct from a browser profile you do not intend to expose. A CDP connection should be treated as access to the attached browser session, so connect only to a browser you control.
Connect WSL2 Hermes to Windows Chrome
When Hermes runs inside WSL2 but the browser with your signed-in session is on Windows, a direct /browser connect can be unreliable. The documented route uses Windows interop to run a stdio MCP bridge from WSL and attach that bridge to Chrome.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems- Add the Windows bridge from the Hermes CLI using the documented command:
hermes mcp add chrome-devtools-win --command cmd.exe --args /c npx -y chrome-devtools-mcp@latest --autoConnect --no-usage-statistics - Test discovery:
hermes mcp test chrome-devtools-winThe test connects to the named server, lists discovered tools, and returns documented status codes indicating success or failure.
- Load the server into a session. Start a fresh Hermes session or run
/reload-mcp, then ask Hermes which MCP-backed tools are available. - Verify browser attachment. If automatic connection fails, check that Windows Chrome is running and reduce background or frozen Chrome tabs; the guide notes that this may help when
autoConnecttimes out.
Windows-mounted paths such as /mnt/c/Users/<you> can avoid UNC current-directory warnings. Replace <you> with the Windows user directory in your environment; it is not a literal path to copy. The connection pattern is Hermes in WSL, a Windows interop stdio MCP bridge, and Windows Chrome. This differs from attaching directly to a browser running inside the same Linux environment.
Configure another MCP server in Hermes
For Browser Use, Browserbase, Firecrawl, or another MCP server, add its entry beneath mcp_servers in ~/.hermes/config.yaml. A local stdio entry uses a command and its arguments—for example, an npx invocation for a Node-based server. A remote HTTP server uses Hermes’s remote-server support. Use the server’s own setup instructions for its exact command, arguments, endpoint, and authentication; these vary by server.
Rank #4
- Add one server under
mcp_serversin~/.hermes/config.yaml, following the server’s documented transport and launch requirements. - Restrict its tools. Use
tools.includeto allowlist necessary actions, ortools.excludeto omit actions you do not want exposed. Scope filesystem access to one project directory and Git access to one repository where applicable. - Turn off unused surfaces. Set
resources: falseandprompts: falseif the server’s resource and prompt wrappers are unnecessary. - Test the server:
hermes mcp test <server>Replace
<server>with the configured server name. Check that the test connects and lists the tools you expect. - Load configuration changes. Run
/reload-mcpafter editing tool allowlists or exclusions, enabled flags, resource or prompt toggles, or credentials. A fresh session also loads the configuration. - Confirm tool availability. In a Hermes chat, ask which MCP-backed tools are available before issuing a browser task.
Keep the initial configuration narrow, especially if a server can access files, repositories, credentials, or a signed-in browser. The Hermes maintainers’ guidance is: “Good MCP usage is not just ‘connect everything.’ It is ‘connect the right thing, with the smallest useful surface.’”
What to expect from browser interaction
Once a browser backend is available, Hermes can use the browser tools discovered from that configuration. In the documented browser model, page content is represented as an accessibility-tree snapshot and interactive elements have reference IDs. This gives the agent a structured way to identify controls. It does not establish that every browser backend exposes identical tools or behaves identically; ask Hermes for the discovered tool list and use the server’s documentation for backend-specific actions.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchBest Value
For a session where the user is already signed in, local CDP is the relevant choice if Hermes and that browser can reach each other. For managed cloud sessions or anti-bot capabilities, consider the documented cloud options. For a Windows browser controlled from WSL2, use the bridge route rather than assuming the local CDP endpoint is reachable across environments.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Troubleshooting
hermes mcp test cannot connect or does not list tools
- Check that the server name in the test command matches the name configured under
mcp_servers. - Check the server’s documented launch command, arguments, transport, endpoint, and credentials. A failed process launch or unreachable remote endpoint prevents tool discovery.
- After changing configuration, reload MCP or start a fresh session before testing from chat. The test command itself connects and lists tools, but a chat session also needs to load its MCP configuration.
Hermes does not show a tool you expected
- Ask Hermes which MCP-backed tools it discovered and compare that list with the server’s available tools.
- Review
tools.includeandtools.exclude; an allowlist that omits an action or an exclusion that names it will hide it. - Check whether the server is enabled and reload after changing enabled flags or tool filters.
/browser connect cannot attach
- Confirm the CLI is running in the environment that can reach the browser’s CDP endpoint. The default is
http://127.0.0.1:9222; use the specific WebSocket endpoint when the browser is exposed elsewhere. - Use
/browser statusto inspect connection state. For WSL2-to-Windows Chrome, use the documentedchrome-devtools-mcpbridge if direct attachment is unreliable. - For a manual browser launch, use a dedicated user-data directory and ensure the browser supports remote debugging.
Windows bridge times out or prints a path warning
- If
autoConnecttimes out, try reducing background or frozen Chrome tabs, as the guide suggests. - If Windows interop emits a UNC current-directory warning, use a Windows-mounted path such as
/mnt/c/Users/<you>to avoid that warning. - Run
hermes mcp test chrome-devtools-winagain, then start a fresh session or reload MCP so the bridge tools are available.
A browser command does not work from WebUI or a messaging gateway
/browser connect, /browser status, and /browser disconnect are interactive CLI commands. Use them in Hermes CLI rather than expecting gateway chats such as WebUI, Telegram, or Discord to dispatch them.
Or skip the browser setup
If the task is to produce a website screenshot rather than interact with a live browser session, ScreenshotNeo is a separate option: a single GET request returns an image or PDF. It does not attach Hermes to Chrome or replace browser automation. For developers who only need the rendered capture, it avoids browser installation and session setup.
Here is the cURL request, with the target URL set to Stripe as in the API example. See the ScreenshotNeo API documentation for parameters and output options.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpScreenshotNeo accepts cookie or consent banners before capture and removes more than 60 known consent platforms, newsletter popups, and chat widgets; each step can be turned off. Bot checks or CAPTCHAs, blank pages, timeouts, failed loads, and cache hits are not billed, and each response indicates the page verdict and billing status in headers. Its MCP server offers take_screenshot, get_page_info, and capture_pdf tools for AI agents, including Claude, Cursor, and other MCP clients. The Free plan includes 1,000 shots a month with no card; paid plans start at $5 for 3,000 shots. Every feature is available on every plan.
Sign up for ScreenshotNeo’s free plan to get 1,000 screenshots a month with no card.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
