vela-mcp

Vela MCP is the MCP server engine for AI-driven dynamic testing of Electron apps.

It exposes 20 tools over the Model Context Protocol, letting an AI agent (Claude Code, Codex, etc.) launch and drive an Electron application, inspect IPC channels, evaluate code in any context, capture screenshots, set breakpoints, record console/network traffic, and save/run re-runnable specs — all without modifying the app's production code.

Installation

npm install -g vela-mcp
# or use npx / a local path

Registering with Claude Code

Add to .mcp.json in your project root (or ~/.claude/mcp.json globally):

{
  "mcpServers": {
    "vela": {
      "command": "node",
      "args": ["<absolute-path-to>/node_modules/vela-mcp/dist/index.js"]
    }
  }
}

Or if installed globally / via npx:

{
  "mcpServers": {
    "vela": {
      "command": "npx",
      "args": ["vela-mcp"]
    }
  }
}

Registering with Codex

Add the same command/args block under the tools.mcp key in your Codex config.

Tools (20 total)

Launch & Targets

| Tool | Description | |------|-------------| | vela_launch | Launch an Electron app and attach CDP | | vela_list_targets | List open renderer/webview/BrowserView targets | | vela_attach_target | Attach to a specific target | | vela_screenshot | Capture a screenshot of the current target |

Eval & IPC

| Tool | Description | |------|-------------| | vela_eval | Evaluate JS in the currently attached renderer | | vela_ipc_list | List registered IPC channels (requires vela-agent) | | vela_ipc_invoke | Invoke an IPC channel and return the result | | vela_main_eval | Evaluate JS in the main process (requires vela-agent) |

Debugger

| Tool | Description | |------|-------------| | vela_debug_attach | Attach the debugger to a target | | vela_set_breakpoint | Set a breakpoint by URL + line number | | vela_wait_paused | Wait until execution hits a breakpoint | | vela_eval_frame | Evaluate an expression in the paused call frame | | vela_step | Step over / in / out | | vela_resume | Resume execution | | vela_debug_detach | Detach the debugger |

Specs

| Tool | Description | |------|-------------| | vela_spec_save | Save a recorded interaction as a re-runnable spec | | vela_spec_run | Run a saved spec and return results | | vela_spec_list | List all saved specs |

Console & Network Capture

| Tool | Description | |------|-------------| | vela_capture_start | Start capturing console logs and/or network requests | | vela_capture_read | Read captured events so far | | vela_capture_stop | Stop capturing and return final buffer |

More

See the repo for the full playbook, Claude Code skill, and sample app.