Before You Begin

All of these come from the normal webSlinger installation. If you have already installed webSlinger and run an automation, you are most of the way there.

You Do Not Need Python

The installer bundles its own Python runtime and every dependency the MCP server needs, including the MCP SDK. There is nothing to pip install and nothing to clone.

Where Things Live

Two directories are involved, and they are deliberately different. Program files are installed under Local AppData; everything that represents your state lives under your user profile.

Reinstalling or upgrading replaces the first directory and leaves the second alone.

Three Steps

1. Create your MCP config file

   Create a file named mcp_config.json in %USERPROFILE%\webSlinger\taskSpinner\. A template named mcp_config.example.json ships with the server under %LOCALAPPDATA%\webSlinger\taskSpinner\; copy it across and fill in your key.

   On a brand-new machine that folder does not exist yet — it is created the first time taskSpinner or the MCP server runs. Create it yourself, or do step 2 first and it will be waiting for you.

   mcp_config.json

   Copy

   {
       "api_key": "wsk_your_webslinger_api_key_here"
   }

   The file must be strict JSON — no comments, no trailing commas. A malformed file stops the server from starting at all.

2. Grant automation permission

   Open the Launch page, start any automation once, and click Enable when Chrome asks. This does two things: it grants webSlinger permission to drive automations, and it registers your Chrome profile and extension so the MCP server knows which browser copy to launch.

   Skipping this produces launch_failed with reason profile_not_registered on the first run.

3. Register the server with your MCP client

   Point your client at the installed wrapper, which invokes the bundled Python for you:

   the path to register

   Copy

   %LOCALAPPDATA%\webSlinger\taskSpinner\mcp_server.bat

   Most MCP clients store this path literally and will not expand %LOCALAPPDATA%, so use the fully written-out form, substituting your Windows username:

   Copy

   C:\Users\YourName\AppData\Local\webSlinger\taskSpinner\mcp_server.bat

   Exact commands for each client are on the Client Configuration page.

Keeping the Key Out of a File

The config file stores your API key in plaintext. If that is a problem — a synced profile, a shared machine, a directory you might commit — set the WEBSLINGER_API_KEY environment variable instead. It overrides api_key in the file, so you can leave that field out, or skip the file entirely: every other setting has a working default.

Verifying It Works

Ask your agent to list your webSlinger scripts. A working server returns the automations on your account. If something is wrong, every tool returns a JSON object with "success": false and a status value naming the problem — check it against the Troubleshooting table.

Claude Code

Run this once in a terminal. --scope user makes the server available in every project rather than just the current directory.

claude mcp add webslinger --scope user -- "C:\Users\YourName\AppData\Local\webSlinger\taskSpinner\mcp_server.bat"

Confirm it registered:

claude mcp list

To supply the API key through the environment instead of the config file:

claude mcp add webslinger --scope user --env WEBSLINGER_API_KEY=wsk_your_key_here -- "C:\Users\YourName\AppData\Local\webSlinger\taskSpinner\mcp_server.bat"

Claude Desktop

Edit %APPDATA%\Claude\claude_desktop_config.json and add a webslinger entry under mcpServers. Restart Claude Desktop afterward.

{
    "mcpServers": {
        "webslinger": {
            "command": "C:\\Users\\YourName\\AppData\\Local\\webSlinger\\taskSpinner\\mcp_server.bat"
        }
    }
}

If you already have other servers configured, add webslinger alongside them rather than replacing the mcpServers block.

Claude Agent SDK

Add an entry to ClaudeAgentOptions.mcp_servers:

"webslinger": {
    "command": r"C:\Users\YourName\AppData\Local\webSlinger\taskSpinner\mcp_server.bat",
    "env": {"WEBSLINGER_API_KEY": "wsk_your_key_here"}
}

Pointing at mcp_server.bat rather than your own interpreter keeps the server on its bundled Python and its own copy of the MCP SDK, so it cannot conflict with the package versions in your project's environment.

The Twelve Tools

Every tool returns a JSON object with a success flag. Expected failures come back as "success": false with a status value the agent can branch on, rather than as an error.

Status Values Worth Knowing

Common Symptoms

Where to Look

Logs are written to %USERPROFILE%\webSlinger\taskSpinner\logs\taskSpinner_YYYYMMDD.log, shared by the native messaging host, the scheduler, and the MCP server.

Per-run records, including the output data and execution report, are kept under %USERPROFILE%\webSlinger\taskSpinner\outputData\mcpRuns\. The 50 most recent are retained by default.

Questions