Skip to content

Teeny Tiny,changelog.

New features and improvements, newest first.

In beta. Talk to us.

  1. The beta, a CLI for scripts, and this changelog

    • Teeny Tiny is in beta, with one self-serve plan: $100 a month for 1,460 machine-hours and up to 2 machines at once. Need more? Talk to us.
    • A new front page shows the ways in, from Agent and the desktop to SSH and Beam, with the CLI’s install command one copy away.
    • Docs are shorter and easier to scan, with a preview of each section.
    • teenytiny help <command> shows focused help with examples, teenytiny help --json describes every command, and --json output stays clean when it’s piped.
    • teenytiny create --idempotency-key <key> makes a retried create return the same machine instead of a second one.
    • With --json, stop, destroy, and ssh keys remove need --yes instead of asking for confirmation.
    • Find text in a terminal, on machines, in Agent, and in Beam, and read a terminal’s output as plain text.
    • Agent can read resources and prompts from your MCP servers, and connect to servers that require DPoP.
    • When a reply runs into the model’s output limit, Agent picks up where it stopped.
    • Removing a passkey now tells your browser or password manager, so it stops offering it.
    • This changelog, and a blog.

  2. Your machines, from ChatGPT

    • Connect Teeny Tiny to ChatGPT or Codex to see your machines, create, start, and stop them, and run commands on the running ones, within your account’s normal access and limits.
    • Connected apps appear under Security, where you can disconnect one at any time.
    • Agent’s controls and icons are more consistent across the workspace.
    • New Support, Privacy, and Terms pages.

  3. Projects with more than one repository

    • An Agent session can work with more than one GitHub repository, and projects are easier to manage from the sidebar.
    • Managed models get more time to stream long tool calls, and a stream that runs out of time now ends with a clear error.
    • Your account’s skill library holds up to 100 skills.

  4. Previews, questions, and scripts in Agent

    • HTML in the agent’s replies renders as a preview you can expand or save.
    • Annotate lines in a file, part of a reply, or an element in the browser to point the agent at exactly what you mean. Browser annotations include a screenshot and the page structure.
    • The agent can ask you a question and wait for your answer, and browser notifications tell you when a session needs you.
    • Save Code Mode scripts and run them again later.
    • Search the transcripts of past sessions.
    • Find any file in the workspace by name, and create, rename, or delete files from the file tree.
    • A clearer model picker.

  5. Drive for every account

    • Drive is available on every account: one place for files at /config/drive on supported machines, in the dashboard, and as a network drive in Finder or File Explorer.
    • Agent can run code in Code Mode, use skills from your account’s library or import them from GitHub, and connect MCP integrations.

  6. Compare

    • A new Compare page explains where Teeny Tiny fits among cloud computers and sandboxes, with a dated market directory.

  7. A polish pass

    • Pages share one layout, keyboard users get a skip link, and signup checks your details before you submit.

  8. Bring your own models

    • Connect your own ChatGPT, OpenCode Go, or OpenRouter account and use its models alongside the managed ones.

  9. Alpha subscriptions and a shared browser

    • Subscribe to the Teeny Tiny Alpha from Pricing: one monthly price for the account, not for each machine.
    • Agent has an integrated browser that you and the agent share, and the agent can drive it.
In beta. Talk to us.

Tiny machines, for your mighty work

Talk to usEmail us