Connector Desktop · Private pilot

Your local AI. Your hardware. Your GPT.to workspace.

GPT.to Connector Desktop links LM Studio, Ollama and compatible local endpoints to GPT.to without exposing a model port to the internet. Your computer opens one encrypted WSS session over outbound port 443.

Choose your platform Review the security model

Outbound onlyNo public LM Studio port and no inbound firewall rule.

Typed capabilityOnly approved local_llm.openai_compatible invocations.

Fails closedPending work remains retryable; there is no silent cloud fallback.

GPT.to Connectorv1.0.0

Outbound only · WSS 443

LOCAL AILM Studio127.0.0.1

DESKTOPConnectordevice identity

CLOUDBrain V2typed invocation

01 provider discovered on loopback 02 exact model preflight passed 03 encrypted device session connected

No inbound port · No LAN route · No silent fallback

Desktop apps

One connector for Windows, macOS and Linux.

Version 1.0.0 · pilot channel. The page identifies your current platform but always leaves every build visible.

Public downloads are intentionally gated.

The cross-platform pilot artifacts have passed packaging and checksum verification, but the Windows and macOS binaries are not yet code-signed/notarized. Download links will activate only after the release-signing gate passes.

Pilot build · public signing gate pending

Windows

Windows 10/11 · 64-bit

NSISx64

gptto-connector-1.0.0-win-x64.exe

SHA-256

be30173558174f6236b3a40a080385d5b32c3dc1246290e09f33c72f9c018ad1

Pending signed release

macOS

macOS · Apple Silicon

DMGApple Silicon

gptto-connector-1.0.0-mac-arm64.dmg

SHA-256

2a06edbedb366b34ce80e80f539849246275d0a8180dce32a7de7391d52db62e

Pending signed release

Linux

64-bit Linux · AppImage or Debian/Ubuntu

AppImagex64

gptto-connector-1.0.0-linux-x86_64.AppImage

SHA-256

c2378485b4af260cbf2919f68cbd8e9e890f5ff944dc1ea66d9e3d34bde200b8

Pending signed release

debamd64

gptto-connector-1.0.0-linux-amd64.deb

SHA-256

b06e81f92f128d8966cdaa15b3d0c63ee4e6e45011f6e66062b02cbfc9f29738

Pending signed release

Private by design

Your model stays local. GPT.to receives only the work you approve.

The connector is a controlled bridge, not a public model server. Device identity, workspace pairing and revocation are enforced independently for every installation.

LM Studio, Ollama and compatible providers remain bound to loopback.

The desktop initiates WSS over port 443; EC2 never calls a LAN address.

Credentials are stored in the operating system secure store and can be revoked per device.

Strict model preflight prevents requests from reaching a different or missing model.

Three steps

Connect a private model without opening your network.

Install the desktop app

Launch Connector Desktop and let it run from the tray. Autostart and reconnection keep the secure session available.

Pair your GPT.to workspace

Sign in through GPT.to and approve one device identity. You can revoke that installation without affecting any other device.

Select and verify the model

Discover LM Studio, Ollama or another compatible loopback endpoint, choose the exact model and pass strict preflight before activation.

Want access to the Connector pilot?

Request private access and we will enable the signed build for your workspace when the distribution gate opens.

Request pilot access