# Hermes Agent

> Let Hermes Agent by Nous Research solve reCAPTCHA image challenges with the Solvecha extension in its browser, or with the Solvecha MCP server.

[Hermes Agent](https://hermes-agent.nousresearch.com) by Nous Research can control a real Chromium browser through the Chrome DevTools Protocol (CDP). Two ways work with Solvecha.

## Way 1: the MCP server (fastest)

Add the server to `~/.hermes/config.yaml`:

```yaml
mcp_servers:
  solvecha:
    url: "https://api.solvecha.com/mcp"
    headers:
      Authorization: "Bearer ${SOLVECHA_API_KEY}"
```

Put `SOLVECHA_API_KEY=slv_...` in `~/.hermes/.env`. Run `/reload-mcp` in a running session. Then ask Hermes to use `solve_recaptcha_grid` when it meets an image challenge. Details and tools: [MCP server](https://solvecha.com/docs/mcp-server.md).

## Way 2: the extension in the browser Hermes drives

The [Solvecha extension](https://solvecha.com/docs/browser-extension.md) reads the challenge, asks the API and clicks the tiles. Hermes does not need to do anything.


1. [Download the extension](https://solvecha.com/download/extension/chrome) and unzip it into a folder, for example `~/solvecha-extension`.
2. Start Chrome or Chromium yourself with a separate profile, the extension and a debugging port. Chrome 136 and later ignore the debugging port on the default profile, so the profile folder is needed.
3. Open the extension options and save your API key once. Turn on auto-solve.
4. In the Hermes CLI, run `/browser connect`. The default address is `http://127.0.0.1:9222`. For another address use `/browser connect ws://HOST:PORT`.
5. Ask Hermes to do the task. When a reCAPTCHA image challenge appears, the extension solves it.


```bash
chromium \
  --user-data-dir="$HOME/.hermes/chrome-solvecha" \
  --remote-debugging-port=9222 \
  --load-extension="$HOME/solvecha-extension" \
  --no-first-run --no-default-browser-check
```

On a server without a screen, run the same command with `xvfb-run`. Extensions need a headed browser.


> **Note:**
> `/browser connect` works in the terminal (`hermes` or `hermes chat`), not in gateway chats such as Telegram or Discord. Chrome 144 and later can ask "Allow remote debugging?" for each connection. A dedicated profile folder avoids most of these prompts. Newer Chrome builds may limit the `--load-extension` flag. Use Chromium or Chrome for Testing if the extension does not load.


## Fewer challenges: CloakBrowser

[CloakBrowser](https://solvecha.com/docs/cloakbrowser.md) is a Chromium build with a changed fingerprint. It aims to keep captchas from appearing. Solvecha solves the ones that still appear. You can start CloakBrowser with a debugging port and the extension and connect Hermes to it in the same way. Read the license facts on the [CloakBrowser page](https://solvecha.com/docs/cloakbrowser.md) first. Solvecha is not affiliated with CloakHQ.

## Add the skill

Hermes reads skills from `~/.hermes/skills/`. Copy the [Solvecha agent skill](https://solvecha.com/docs/agent-skill.md) there so that Hermes knows when to use which way.