October 21, 2026

Transit is now free and open source.

The whole desktop client — SSH, serial, SFTP, packet capture, the API client, the local terminal, and the investigation-only AI agent with its policy gate and approval modal — is published as Transit Community Edition under the Apache License 2.0. No account, no subscription, no telemetry. The AI runs on your own Anthropic or OpenAI key, or on models on your own hardware.

One-line install. Each script shows its plan and asks before installing, and verifies the published checksum first. macOS and Linux (Debian/Ubuntu, Fedora, Arch and derivatives; AppImage elsewhere):

curl -fsSL https://raw.githubusercontent.com/transit-community/transit/main/install.sh | sh

Windows (PowerShell 7, no administrator prompt). The built-in Windows PowerShell 5.1 is not supported: install PowerShell 7 with winget install Microsoft.PowerShell first, or use the setup.exe from the release page.

irm https://raw.githubusercontent.com/transit-community/transit/main/install.ps1 | iex

Download for macOS, Windows, or Linux View the source on GitHub

If you have a subscription

You were emailed the same details. In short:

Monthly plans keep working until your current renewal date and are not charged again. Nothing to cancel.
Annual plans keep working until November 5, 2026, and a prorated refund for the unused months is issued that day to the original payment method. Card refunds take five to ten business days to show.
The hosted AI service shuts down on November 17, 2026. After that the installed Transit app still works as a terminal, but its cloud AI will not. Install the Community Edition and add your own key before then. On first launch it imports your devices and host-key pins from the Transit app, and device passwords in your OS keychain carry over; only the AI provider key is new.
Account data is deleted after the shutdown in line with the Privacy Policy. Sign in to your account any time before then for invoices.

How the AI runs now

Settings → AI providers takes an Anthropic or OpenAI API key, stored in your OS keychain and sent only to that provider's own API. You pay the provider directly, per token. The model list is fetched live from your account. Prefer no key? Settings → Local models runs the assistant on Ollama or any OpenAI-compatible server on your own hardware.

Step-by-step key setup for both providers is in the README under Get an API key. The providers' own guides, both billed per token (set a spend limit while you are there):

Anthropic: Get started · API keys
OpenAI: Developer quickstart · API keys

Preview: Sign in with ChatGPT. Instead of an OpenAI key, the OpenAI side can run on a personal ChatGPT Plus or Pro plan through OpenAI's sign-in for open-source apps. Choose it on the OpenAI card under Settings → AI providers; the browser opens, and usage counts against your plan. Set a weekly cap for the app under ChatGPT → Settings → Usage. Business, Enterprise and Edu workspace accounts are not eligible, and OpenAI controls the preview and may change it. Anthropic does not permit consumer plans in third-party apps, so the Anthropic side is API-key only.

The rules have not changed: the agent is read-only, every command it proposes passes the policy gate and your explicit click, and the binary contacts nothing you did not configure — now checkable in the source.

Contributing

Bug reports, vendor policies, and features are welcome through GitHub. Open an issue before a pull request; the review checklist is in CONTRIBUTING.md. Security bugs go through GitHub's private vulnerability reporting, never a public issue.