get_screenshotLABSCO SUMMARY
It closes the loop between writing a test and watching it fail.
Most test-authoring assistants can write steps but cannot see the result, so a broken selector comes back as a guess. Here the same connection that adds commands also runs the test and returns per-step results and annotated page HTML, which is what makes 'run it and fix the failures' a real instruction. The dependency is absolute, though: with Selenix closed or its MCP switch off, every call fails at the bridge.
What it is
A bridge between an MCP client and a running Selenix desktop install. The MCP server translates tool calls into HTTP requests to a bridge inside the app, so a model can read the current test, add or fix commands, run it, and look at the browser while it does.
What you get
- Reading what exists —
get_test_list,get_current_test,get_test_by_name,get_project_infoandget_active_suite - Seeing the page as the test sees it —
get_screenshotcaptures the browser window,get_page_htmlreturns the HTML with interactive elements annotated,get_search_htmlsearches that HTML by regex - Command reference on demand —
get_command_listenumerates every Selenix command andget_command_inforeturns the documentation for one, so the model does not have to guess syntax - Writing tests —
add_commandsinserts at a position,clear_and_replace_commandsrewrites a test,fix_commandsapplies targeted updates, removals or insertions run_testexecutes a test and waits for the result, up to 2 minutes- Reading the outcome —
get_step_resultsreturns pass and fail per step,get_logspages through recent logs 20 at a time,get_workspace_contextsummarises the workspace
Requirements
The Selenix desktop app installed and running, with System Settings > MCP Server set to Yes — that switch starts the local bridge the MCP server talks to. Also Node.js 18 or later. Install the npm package @selenix/mcp-server (version 1.0.7) globally, or point the client at the copy bundled inside the Selenix install. The bridge listens only on 127.0.0.1 and authenticates with a bearer token regenerated each time Selenix starts; connection details live in ~/.selenix/bridge.json and are re-read on every call, so restarting Selenix does not mean restarting the client.
Setup effort
Follow the published steps — npm install -g @selenix/mcp-server
USE CASES
Building a test from a description
Describe the flow, let the model look up the command syntax, and have it write the steps into the test.
Fixing a test that broke on a selector
Run it, read the failing step, look at the annotated page HTML, and patch just that command.
Finding your way around an inherited project
List the tests and the active suite, then read the one you care about.
Seeing what the browser is actually showing
Capture the window mid-run instead of inferring the page state from the log.
18 tools
Listed in project docs. Your client may show a different list.
get_page_htmlget_search_htmlget_selected_command_infoget_current_testget_test_listget_test_by_nameget_command_listget_command_infoget_project_infoget_active_suiteget_step_resultsget_logsget_workspace_contextadd_commandsrun_testclear_and_replace_commandsfix_commandsBEFORE YOU INSTALL
What to know first
Tools you get
18 tools named in the project's documentation.
Requires
node on PATH
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
In Selenix: System Settings > MCP Server > set Enable MCP Server to Yes
npm install -g @selenix/mcp-server
npm root -gAdd the Claude Desktop block to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json; Windows: %APPDATA%\Claude\claude_desktop_config.json), then restart Claude Desktop
MCP Clients
Claude Desktop
{
"mcpServers": {
"selenix": {
"command": "node",
"args": [
"<YOUR_NPM_ROOT_PATH>/@selenix/mcp-server/dist/bundle.js"
]
}
}
}