One bot
became many.
Give each task its own persona with its own model, memory and workspace — and run every one of them from a single chat window. Behind that: GPT-6 Astra, a real Linux desktop in your browser, and customers who finish their own setup.
One bot, many minds
Personas are live on running OpenClaw instances. Create a dedicated assistant for writing, for research, for whatever you keep re-explaining — each with its own model, and its own memory, skills and workspace kept separate from the others.
On Hermes the + button in chat names and creates another bot inside the same instance, with no second deploy. Quick Chat now lists every bot in a sidebar with its name, model and online status, so you switch between them without leaving the window — and a Hermes bot can now message another Hermes bot.
GPT-6 Astra, on your key or your subscription
Choose GPT-6 Astra with your own OpenAI key on OpenClaw, or through a connected ChatGPT subscription on OpenClaw and on Hermes when your account has access to it. GPT Image 2.5 Flare and Sunburst arrive for image generation. New paid bots now start on GLM 5.3 Flash, and Kimi K3, Gemini 3.7 Flash, Grok 4.6 and Agnes 3.0 Flash join the picker.
Open your bot's desktop
Your bot has a Linux desktop, and now you can use it. Open Live Desktop from the bot's card and its browser, terminal and file manager are right there in a browser tab. Or connect the bot to your own Tailscale network and reach it over SSH from your own machine.
Your customers finish their own setup
Customers of a bot you resell can connect WeChat, Telegram, WhatsApp, Discord or Lark themselves, from inside your storefront — and enter their own model key in the same place. Eight providers are accepted now, including Anthropic, Moonshot, xAI and OpenCode Go. Your storefront gained a password reset link, and a starter credit lets you deploy a customer bot before adding a card.
Seven new tools, including Blender
Point your bot at a Blender session running on your own machine and it can read the scene, edit objects and materials, run Python, and pull assets from Poly Haven, Sketchfab and Poly Pizza. Joining it: VoiceStudio, SkillSpector, AiToEarn for scheduling social posts, ai-memory, Free Claude Code and FreeLLMAPI — plus Vast.ai GPU control from chat.
Also shipped
Link WhatsApp to Hermes with a phone number and a pairing code, so setup finishes on one phone.
A skill created or edited on one bot reaches your other bots that have automatic skill sync enabled, without waiting for the hourly refresh.
Files your bot produces stay attached even if it sends another message afterward.
Videos generated in dashboard chat arrive in an inline player when the render finishes.
Backup downloads use your browser's download manager and survive longer exports.
After an overdue subscription is paid, affected bots resume automatically.
Recent Hermes messages appear sooner, and the chat dock stays visible while the panel loads.
Updating a provider key keeps the models you selected, custom and fallback included.
Watch a coding workspace's run output as it arrives, through the Developer API.
A newly created website can no longer show another site's content.
Agnes Video 2.5 Flash for video generation, including on the free trial.
New OpenClaw bots come ready to install Python packages for scripts you ask them to run.
All in your dashboard.
Deploy an OpenClaw or Hermes instance and everything above is one click away.