Shogo 2.0
Shogo 2.0 introduces personal workspaces, a free space where your own AI companion works alongside you, and Hoshi 2.0, our new default model. It also rolls up everything shipped since 1.12: a redesigned workspace, scheduled routines, new models, and a long list of reliability fixes.
Personal workspaces
- A free personal space for everyone. Every account now includes a personal workspace alongside a team workspace, so you get two free workspaces: one just for you and one to share.
- Your own AI companion. Your personal space comes with a companion you can name and give an avatar and personality. It runs on Hoshi 2.0 by default.
- Goals with approvals. Give your companion a goal and follow along as it plans, reports progress, and asks for your approval before important steps. Results show up as artifact cards you can open directly.
- Scheduled routines. Ask your companion to do something on a schedule, for example "every weekday at 9am, summarize my inbox", and it runs automatically in your timezone. Ask your companion to change, pause, or remove a routine at any time.
- Side chats and notes. Start side conversations without losing your main thread, and keep persistent notes your companion can work with.
- See what's happening now. The Activity page shows chats and agent tasks that are running right now, in both personal and team workspaces.
Hoshi 2.0
- A new default model. Hoshi 2.0 now powers your personal companion and Shogo's built-in assistant by default. It is backed by DeepSeek-V4.1-Flash with thinking enabled, so it works through a problem before it answers.
- Same price. Hoshi 2.0 costs the same as Hoshi 1.0.
- Available on the API. Use
hoshi-2-0with the OpenAI-compatible public API. Hoshi 1.0 (hoshi-1-0) stays available as a legacy model while you migrate.
New
- Destination-first onboarding. New accounts start by choosing whether they're setting up for themselves or a team, can invite teammates during setup, and get a short checklist to get going.
- Redesigned workspace navigation. A unified chat shell, workspace tabs that match across web and desktop, a project sidebar you can open from chat, and a native "liquid glass" look on iOS.
- Chat and plan history search. Find past conversations and plans quickly.
- Universal chat dock. Queued messages, running tasks, and checklists live in one dock above the composer, and each turn's agent work collapses into a tidy "Worked for X" summary.
- Survives connection drops. If your internet drops mid-turn, chats pause and pick up where they left off, and messages you send while offline are queued and delivered when you're back.
- Audio input. Models that understand audio can now take voice and audio input directly.
- Docker Compose projects. Run multi-service projects on Shogo with exposed ports, tunnels, and public previews.
- Shogo in Slack. Install the Slack app and talk to one Shogo agent for your whole workspace. It pulls in the right project as the conversation needs it.
- Custom subagents. Drop agent definitions into
.shogo/agents/*.mdand they load automatically. - New models. Claude Opus 5, Sonnet 5, and Fable 5.1, plus GPT-6 Astra and the GPT-5.6 family (Sol, Terra, Luna). Image generation now defaults to gpt-image-2.5-flare.
- Desktop beta channel. Opt in under Settings → Updates to get new desktop builds as soon as they land on main.
- Publishing upgrades. Push your dev database to the live published app, keep static apps always-on, and show a friendly loading page while a sleeping app wakes up.
- New templates. Restaurant website and storefront templates, plus support for your own custom routes.
- Appearance settings. Customize how Shogo looks, with a redesigned process panel.
- What's New. You'll now see a short summary like this one in the app after each major release.
Improved
- Context usage breakdown. Click the context meter in chat to see what's using your context window.
- Model picker. Hover a model to see more detail before you pick it.
- Mobile polish. A smoother project drawer, better keyboard handling in search, cleaner activity and marketplace screens, and a unified color palette across mobile and web.
- Canvas builds. Canvases are type-checked before they're shown, and build errors appear in the build log instead of failing silently.
Fixed
- Chat reliability. Fixed slowdowns while responses stream in, and assistant messages not saving in workspace chats without a project.
- Images. Retired DALL-E models were removed, and images from your personal companion now display correctly.
- Windows desktop. Hardened chat, process handling, and cloud recovery on Windows, and fixed path handling for beta updates.
- Live models. Desktops connected to the cloud now see Live models correctly.
Questions or feedback? Reach us through the in-app Help menu.