Short answer: VenelX Studio includes a local Model Context Protocol (MCP) server. Connect it to Claude Code, Codex, Cursor or another MCP client, and the agent can list your projects, start builds, boot and control simulators, create test-build links and run store releases on your Mac, using about 60 tools. A read-only mode exposes only the tools that look and never change anything.
What MCP adds to mobile work
AI coding tools are good at editing code and weak at everything after it: they can't see whether the build passed, can't open the app in a simulator and can't tell you which build is in TestFlight. MCP is the standard way to give a tool those abilities. A server describes a set of tools, and the agent calls them.
The Studio server runs on the same Mac as your projects and talks to the Studio app over a local socket that only your user account can open. It does not listen on the network.
What the agent can do
| Area | Examples |
|---|---|
| Mac and projects | Status, settings, installed tools, storage, list and add projects |
| Builds | Start a build, read its log, see the queue |
| Simulators and devices | List, boot, shut down, install a build, open a deep link |
| Sharing | Create an install link for a build, or a preview link for a dev server |
| Releases | See what is on each store track, upload, promote |
Each tool carries a hint saying whether it only reads, makes a change that can be undone, or does something that can't be taken back, such as shipping a release. MCP clients use those hints to decide when to ask you first.
Connecting it
The simple way: in the Mac app, open Settings › AI tools (MCP) › Connect, and press Connect next to the tool you use. Studio adds an entry named venelx to that tool's configuration, keeps everything else in the file and saves a backup first.
By hand, for Claude Code:
claude mcp add --transport stdio --scope user venelx -- "/Applications/VenelX Studio.app/Contents/MacOS/venelx-engine" mcp
For tools that read a JSON configuration, such as Cursor and Claude Desktop:
{
"mcpServers": {
"venelx": {
"command": "/Applications/VenelX Studio.app/Contents/MacOS/venelx-engine",
"args": ["mcp"]
}
}
}
Then ask the agent something it could not have answered before, such as "which simulators are running?" or "build the iOS app and tell me if it fails".
Keeping it read-only
Add --read-only after mcp, or flip the Read-only switch in the app before connecting. The agent can then inspect projects, builds, logs and release status, and cannot start, change or delete anything. It is a good default while you learn what the agent tends to do.
Pausing remote access from the Mac's menu bar also blocks the tools that reach outside the Mac.
A sensible way to use it
- Let the agent build and read logs freely. That closes the loop on "does it compile?".
- Let it drive a simulator to check its own UI change.
- Keep a person on store releases. Uploading to an internal track is low risk; promoting to production deserves a human, and Studio can require a teammate's approval for it.
Signing keys and store credentials stay in the Mac's Keychain throughout. The agent asks Studio to perform a release; it never receives the key. More about that model is on the security page, and the wider product is described in how it works.