Selenium MCP Server

Instructions

Implementation Reference

• src/tools/actionTools.ts:203-228 (registration)

  Registers the 'browser_key_press' MCP tool with server.tool(), including input schema for 'key' parameter and the handler function that retrieves the WebDriver, instantiates ActionService, calls pressKey(key), and returns success/error messages.

  server.tool(
    'browser_key_press',
    'Press a key on the keyboard',
    {
      key: z.string().describe("Key to press (e.g., 'Enter', 'Tab', 'a', etc.)"),
    },
    async ({ key }) => {
      try {
        const driver = stateManager.getDriver();
        const actionService = new ActionService(driver);
        await actionService.pressKey(key);
        return {
          content: [{ type: 'text', text: `Key '${key}' pressed` }],
        };
      } catch (e) {
        return {
          content: [
            {
              type: 'text',
              text: `Error pressing key: ${(e as Error).message}`,
            },
          ],
        };
      }
    }
  );

• src/services/actionService.ts:60-63 (handler)

  Implements the core key press functionality in ActionService using Selenium WebDriver's ActionSequence: keyDown(key) followed by keyUp(key).

  async pressKey(key: string): Promise<void> {
    const actions = this.driver.actions({ bridge: true });
    await actions.keyDown(key).keyUp(key).perform();
  }

• src/tools/actionTools.ts:206-208 (schema)

  Zod schema defining the input parameter 'key' as a string for the browser_key_press tool.

  {
    key: z.string().describe("Key to press (e.g., 'Enter', 'Tab', 'a', etc.)"),
  },