This document will guide you through setting up and running the Tasker MCP integration, including instructions for installing dependencies, preparing servers, and updating tasks.

• Import dist/mcp_server.prj.xml into your Tasker app.
• After importing, run the MCP generate_api_key task to generate an API key for secure access.

CLI Server:

• From the dist/ folder, select the correct CLI server binary for your device's architecture, such as tasker-mcp-server-cli-aarch64.
• Copy both the binary and the toolDescriptions.json file to your device (phone or PC).
• Rename the binary to mcp-server after copying.

Example:

Using scp:

scp dist/tasker-mcp-server-cli-aarch64 user@phone_ip:/data/data/com.termux/files/home/mcp-server

Using adb push:

adb push dist/tasker-mcp-server-cli-aarch64 /data/data/com.termux/files/home/mcp-server

• Run the server in SSE mode with:

./mcp-server --tools /path/to/toolDescriptions.json --tasker-api-key=tk_... --mode sse

• Or call it through the stdio transport:

payload='{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "tasker_flash_text", "arguments": { "text": "Hi" }  } }'
echo $payload | ./mcp-server --tools /path/to/toolDescriptions.json --tasker-api-key=tk_...

The tasker-mcp-server-cli application accepts the following flags:

• --tools: Path to JSON file with Tasker tool definitions.
• --host: Host address to listen on for SSE server (default: 0.0.0.0).
• --port: Port to listen on for SSE server (default: 8000).
• --mode: Transport mode: sse, or stdio (default: stdio).
• --tasker-host: Tasker server host (default: 0.0.0.0).
• --tasker-port: Tasker server port (default: 1821).
• --tasker-api-key: The Tasker API Key.

• Connect your MCP-enabled application by pointing it to the running server.

{
  "mcpServers": {
    "tasker": {
      "command": "/home/luis/tasker-mcp/dist/tasker-mcp-server-cli-x86_64",
      "args": [
        "--tools",
        "/home/luis/tasker-mcp/dist/toolDescriptions.json",
        "--tasker-host",
        "192.168.1.123",
        "--tasker-api-key",
        "tk_...",
        "--mode",
        "stdio"
      ]
    }
  }
}

• Install Go using your package manager:

sudo apt-get install golang-go

• Build the CLI server (cross-compiling example for ARM64):

cd cli
GOOS=linux GOARCH=arm64 go build -o dist/tasker-mcp-server-cli-aarch64 main.go

Due to limitations in Tasker's argument handling, follow these steps carefully to mark tasks as MCP-enabled:

• Add a comment directly in the task settings. This comment becomes the tool description.

Tasker supports only two positional arguments (par1, par2). To work around this, we'll use Task Variables:

• A TaskVariable becomes an MCP argument if:
  1. Configure on Import: unchecked
  2. Immutable: true
  3. Value: empty

After setting the above values you can also set some additional metadata:

• Metadata mapping:
  • Type: Derived from Task Variable's type (number, string, onoff, etc).
  • Description: Set via the variable's Prompt field.
  • Required: If the Same as Value field is checked.

Note: Temporarily enable "Configure on Import" to set the Prompt description if hidden, then disable it again. The prompt will survive.\

These steps will make sure valid tool descriptions can be generated when we export our custom project later.
Task Variables cannot be pass-through from other tasks, though, so we need to do one last thing in order to get all the variables from the MCP request properly set.

Copy the action MCP#parse_args to the top of your MCP task to enable argument parsing. You can get this from any of the default tasks. But do not modify this action!

Now your custom tasks are ready:

• Export your mcp-server project and save it on your PC.
• Ensure Node.js is installed, then run:

cd utils
npm install
node xml-to-tools.js /path/to/your/exported/mcp_server.prj.xml > toolDescriptions.json

Use this toolDescriptions.json file with your server.

Happy automation!