Integrated native MCP server

Is there an existing issue for this feature request?

• I have searched the existing issues

Is your feature request related to a problem?

I would like to have my AI agent connect to my slicer to read printer, filament, and process settings to help me debug print issues or catch settings issues before I waste time and filament printing.

Which printers will be beneficial to this feature?

All

Describe the solution you'd like

An integrated MCP server that:

• Is disabled by default
• Is read-only by default
• binds loopback only
• grabs a random port
• shows a UI indicator when enabled
• exposed only as an advanced setting

1. On first “Enable MCP”:
   • Generate random high port, e.g. 49152–65535
   • Generate 128/256-bit token
   • Store them persistently:
     • port in Orca config
     • token in Orca user config with restrictive perms
2. On each Orca startup:
   • Bind 127.0.0.1:<stored_port>
   • Require Authorization: Bearer <stored_token>
   • If port is occupied, show “MCP port unavailable” and let user regenerate; don’t silently change it because that breaks the agent config.
3. User configures agent once:

  {
    "url": "http://127.0.0.1:53842/mcp",
    "headers": {
      "Authorization": "Bearer <persistent-token>"
    }
  }

Ideally this work would be split into phases to keep PRs sane:

1. Persistent endpoint/security

• Enable flag
• Stored port/token
• Loopback bind
• Status UI
• Preferences UI
• logs
• auth failure behavior
• integration tests

1. Read-only MCP

• Initialize
• tools/list
• tools/call
• expose selected presets
• expose full config
• expose option metadata/search

1. Writes to active presets

• validate keys and values
• report before/after errors
• apply values via wx main thread

1. Preset selection

• ability to select machine/filament/process presets

Credit to @Wdits in #13451 for first surfacing this request.

Describe alternatives you've considered

1. Agent reads screen via computer use
2. Agent investigates Orca thread memory

Both of these solutions are token intensive and shouldn't be seen as agent native.

Additional context

No response