Ganado Bridge Local

One small connection.One real first task.

Connect Bridge to a local AI app or the ChatGPT Developer Mode preview. Personal and commercial use are free. Start with a disposable test file.

You’ll need a Mac and a compatible local AI app, or access to ChatGPT Developer Mode. Bridge is a tool connection, not a separate chatbot. The hosted connection is a technical preview; it is not approved or published in the ChatGPT directory. Understand the access first.

CHATGPT DEVELOPER MODE

Pair your Mac to ChatGPT

Use the hosted technical preview with your own Mac. Download and extract the bundle, start Bridge, then authorize the connection with its short-lived code.

Follow the ChatGPT setup

Requires Developer Mode access. The directory listing is not approved.

MY APP SUPPORTS LOCAL EXTENSIONS

Download the bundle

Use the packaged .mcpb extension in a desktop AI app that supports it. The dependencies are included.

Download 0.2 preview

Then follow the bundle setup ↓

I USE CURSOR OR ANOTHER LOCAL AI APP

Use a local connection

Extract the same bundle and add a configuration to your app. This route requires a supported Node installation.

Open the Cursor guide

Setup for other local MCP clients ↓

Compatibility: macOS; Node.js 22.22+ within 22.x, or 24.x. An MCPB host supplies the runtime; a manual connection uses your installed Node. This is not a notarized native Mac application. Archive-level tests have passed, but every desktop app and version has not been verified. Follow your app’s normal extension rules—do not disable security controls.

Connect through ChatGPT Developer Mode

1. Download 0.2.0-preview.0 above and extract the .mcpb as a ZIP into a folder you control. Use Node.js 22.22+ within 22.x, or 24.x; the bundle already includes its dependencies.
2. Open Terminal in that folder and run:

   node bin/bridge.mjs connect
   

   Keep Terminal open. Bridge displays a pairing code that expires after 10 minutes.
3. In ChatGPT Developer Mode, create a connection using https://ganado-bridge.vercel.app/mcp as the MCP URL and OAuth authentication. This requires Developer Mode access on your account.
4. Enter the pairing code on the Ganado Bridge authorization page and review the access before confirming. Enter the code in that page, not in the chat.
5. Select the connection in ChatGPT and run the first-task prompt below. Keep your Mac awake and Bridge running.

Technical preview: a working connection is separate from directory publication. Ganado Bridge is not approved or published in the ChatGPT directory.

1. Add the bundle to your AI app

Download the versioned file above, then open or import it in a desktop client that supports local .mcpb extensions. The bundle route does not need a separate npm, Homebrew or ripgrep installation.

Review the publisher and tools. Bridge starts disabled. Enable the setting Enable owner-level local file and terminal access only after reading the MIT licence and access model.

Bridge runs as your OS user and is not a sandbox. Commands can change or delete files and access the network. Requested file contents and results go to your AI provider. No remote SSH host is enabled by default, and this bundle does not connect you to Ganado’s computers.

Using a manual connection instead?

Extract the .mcpb as a ZIP into a folder you control. With a supported Node installed, run these commands from that folder:

node bin/bridge.mjs doctor
node bin/bridge.mjs selftest
node bin/bridge.mjs config

doctor checks the environment. selftest checks disposable fixtures. config prints the connection settings with your own file paths. Review and add those settings to your AI client. It does not edit client settings for you. Passing the diagnostic does not mean your AI app has completed a task.

Cursor setup guide · Fix a disconnected server

Try a real task, safely

Once connected, paste this prompt into your AI app. It creates a disposable test file, makes one checked change and verifies a harmless command.

Use Ganado Bridge to confirm the connected system.
Create a uniquely named folder under the OS temporary directory.
Inside it, create a text file containing BRIDGE_BEFORE.
Read it, use its current SHA-256 to replace that marker with BRIDGE_AFTER,
and read the file back to verify the change.
Run a harmless printf command in that folder and wait for completion.
Report the actual exit code and whether each step was verified.
Do not read or modify anything outside this disposable test folder.

Check the resulting file and command result yourself. A running command is not finished, and a successful edit is not proof that a whole project works. Share an optional first-task result or email us if a step fails. Share only a redacted result, not private paths, credentials or customer data.

Verify the download and release

Current build: 0.2.0-preview.0. Compare the downloaded file’s SHA-256 with this value and the release’s SHA256SUMS:

ce45aed48044e6f926658da621bb763b6dc898ec36a81001d7654cdcd9fd0429

Release notes, checksums and test evidence. A matching archive and automated tests do not certify every client’s installation interface.

Disconnect or remove Bridge

For the hosted preview, stop the running connection with Ctrl-C and run node bin/bridge.mjs disconnect from the extracted folder. Wait for confirmed revocation before deleting local state; if it fails, keep the state and retry. For a local connection, disable or remove Bridge in your AI app and stop its process. Remove the extracted bundle folder when you no longer need it. There is no automatic startup service, global package or LaunchAgent installed by the preview.

Metadata-only audit logs may remain in ~/.local/state/ganado-bridge-agent; remove those separately if not needed. Sessions end when the agent stops. No hidden agent telemetry or licence-server call is included.

Works for you? Keep using it.

MIT includes personal and commercial use. No payment or activation key.

Read the MIT licence

Only use computers and data you are authorized to operate. Both connection routes are previews; contact support if your first task fails.