On this page

Connect your MCP client

Actuator detects your platform, generates the right JSON and writes it into the right file. You should not need to hand-edit anything.

One command per client

Pick yours, run it, then restart the client so it picks up the change.

Setup command for each supported MCP client
ClientCommand
Claude Codedrengr setup --client claude-code --write
Claude Desktopdrengr setup --client claude-desktop --write
Cursordrengr setup --client cursor --write
Windsurfdrengr setup --client windsurf --write
VS Codedrengr setup --client vscode --write
Antigravitydrengr setup --client antigravity --write
Xcodedrengr setup --client xcode --write

Run drengr setup with no flags to print the config instead of writing it, which is what you want if you keep your client config in version control.

Claude Code has a shorter path

Claude Code can add the server itself, without Actuator writing any file:

bash
claude mcp add drengr -- npx -y drengr mcp

From the MCP registry

Clients with registry support can discover Actuator directly. Search for Drengr in your client's server browser, or use the registry identifier:

text
dev.drengr/server

Android Studio

Android Studio cannot spawn a local MCP server, so it connects over HTTP instead. This is the one client where you keep a process running yourself.

bash
drengr setup --client android-studio --write
drengr mcp --http

The HTTP server binds 127.0.0.1 only, on port 7878 by default. If you change it, the port has to match on both sides:

bash
drengr setup --client android-studio --write --port 7900
drengr mcp --http --port 7900

After restarting Studio, Settings then Tools then AI then MCP Servers lists it. Type /mcp in the Gemini panel to see the three tools.

Writing the config yourself

Every stdio client takes the same shape. This is what --write produces:

claude_desktop_config.jsonjson
{
  "mcpServers": {
    "drengr": {
      "command": "drengr",
      "args": ["mcp"]
    }
  }
}

Pin a device with ["mcp", "--device", "emulator-5554"] when more than one is connected. Without it, Actuator auto-detects.

When the tools do not appear

In roughly the order these actually happen:

Common reasons the Drengr tools do not show up
SymptomFix
Config written, tools still missingThe client only reads its config at launch. Fully quit and reopen it, not just the window.
Tools listed but every call failsUsually no device. Run drengr doctor outside the client.
ADB not found, though your shell finds itMCP clients launch servers with a minimal environment. See device setup.
A stale server is answeringdrengr restart from a separate terminal. Your client spawns a fresh one on its next request.
iOS tools fail on first usedrengr build-runner once, from a real terminal. xcodebuild needs a TTY, so it cannot run inside the client.

Server logs are at ~/.drengr/mcp.log. Actuator redirects its own stdout to stderr so it never corrupts the MCP transport, which means the log is the place to look rather than the client console.