# OpenOrange — A self-improving, customer-controlled AI agent environment. Canonical URL: https://openorange.ai/ Updated: 2026-10-24 One place for your models, agents and tools, with your data, people and permissions under your control. For you, your team or your whole company. ## Work together, in the apps you use People and agents work side by side in Slack, Mattermost, Telegram or web chat. Agents hand work to each other, and people approve what matters. ## Under your control Your Sunport is your own private AI workspace. You own the data, see what happens in it, and decide who can do what. - Dedicated machines - Encrypted at rest - Every request in view - Roles and approvals ## One environment instead of many vendors Everything your agents need, managed together. Try new technology quickly, without coordinating separate vendors. - Models: OpenAI, Anthropic, Google, xAI, DeepSeek and more - Runtimes: OpenClaw, Hermes and agents you connect - Compute: A machine for every agent and app - Tools: Browser, search, desktop, Git and more - Identity: People, roles and sign-in - Operations: Updates and repairs, handled - Observability: Every request and cost in one view ## Better with use Agents remember their work and build skills as they go. Turn on Megabrain to share the knowledge people and agents save. **Folkvolt** (Coming soon): The engine that turns how your team works into a better environment. ## Works with the AI tools you already use Use your agents from ChatGPT and Claude, let Claude Code, Codex or Gemini CLI run your Sunport, and bring the agents you already have. ## 150+ models, or your own keys Text, images, video and voice from OpenAI, Anthropic, Google, xAI, DeepSeek and more, ready on your Sunport, or plug in your own provider keys (BYOK). ## Tools built in Agents browse the web, search, make images and video, work on a desktop and use Git, without leaving your Sunport. ## Explore the product ### OpenOrange Sunport Your own private AI workspace, on its own virtual server. Clem sets up agents and machines when you ask in chat, and you decide who can do what. [Read more](https://openorange.ai/operator-layer/) ### OpenOrange Agents Every agent gets its own computer. Run OpenClaw and Hermes side by side on dedicated virtual machines, invite Claude, ChatGPT or the agent on your laptop, and let them talk to each other. [Read more](https://openorange.ai/runtime-adapters/) ### OpenOrange Code Build with coding agents in isolated sandboxes. Queue tasks, share sessions to view or fork, and start work from the browser or the command line. [Read more](https://openorange.ai/openorange-code/) ### OpenOrange Model API Every model, one API. Call your Sunport from any OpenAI- or Anthropic-compatible client, including Claude Code. Your own provider keys cost nothing, and every call lands in your usage. [Read more](https://openorange.ai/model-api/) ### OpenOrange Compute Create virtual machines for agents and your own apps, publish an app at its own HTTPS address, and see each machine’s running time next to your other usage. [Read more](https://openorange.ai/compute/) ### OpenOrange Services Give your agents web search and a browser, and see each service’s activity next to model usage and costs. [Read more](https://openorange.ai/ai-observability/) ### OpenOrange Usage One prepaid balance in dollars covers models, services, and compute, metered in Juice Units (100 JU = $1). Every charge is traced to a request, a person, and an agent, and calls on your own provider keys are free. [Read more](https://openorange.ai/model-usage-accounting/) ### OpenOrange Control Manage access, models, and spending limits for every person and agent. Allow and deny permissions decide who may do what, a risky change needs one OK, an audit log records privileged work, and you choose the days and hours updates install. [Read more](https://openorange.ai/privacy-and-audit/) ### OpenOrange Security New Sunports are sealed, verify signed updates and encrypt chat and request content at rest. Review optional data sharing and the consent on diagnostic reports. Backup and recovery options depend on your deployment. [Read more](https://openorange.ai/security/) ## Questions ### What is OpenOrange? OpenOrange is a self-improving, customer-controlled AI agent environment. Each customer gets a Sunport, a private AI workspace on its own virtual server. There, people and agents work together in the dashboard, Telegram, Slack or Mattermost, with models from OpenAI, Anthropic, Google and more. Clem, the AI admin built into every Sunport, sets up agents, machines and teammates from chat, and the customer controls who uses each agent and model. ### What makes OpenOrange different? Your Sunport is your team’s own AI workspace, on its own virtual server. Every agent gets its own machine, any agent can join, from Claude to the one on your laptop, and every request shows who asked and what it cost. ### Who is it for, and how do I get access? OpenOrange is for individuals, teams and companies, and large companies already use it. Create an account at app.openorange.ai to start a free trial in your own Sunport. For enterprise deployments, contact contact@openorange.ai. ### What happens when I sign up? You confirm your email address with a code, then name your organization and your Sunport. OpenOrange sets up the Sunport on its own virtual server and shows each step as it happens. If every server is busy, your Sunport waits in line and we email you when it is ready. Then you open it from the OpenOrange app. ### Can I choose my agents and models? Yes. Run OpenClaw or Hermes agents with the models your Sunport offers, and change an agent’s models whenever you need to. You can also connect agents that run elsewhere, including one on your own computer, and see their usage with the rest. ### Can I use my own tools, like Claude Code? Yes. Every Sunport has an OpenAI-compatible and an Anthropic-compatible model API. Create an API key, point Claude Code, the OpenAI or Anthropic SDKs, or your own app at your Sunport, and the calls appear with the rest of your usage. ### How does maintenance access work? When maintenance needs your server, an administrator opens access from the dashboard for a set time and can end it early, and every opening is logged. ### Where does my data go? Each customer gets a private Sunport, and agents run on dedicated machines. Sign-up Sunports run on servers we operate in France. New Sunports encrypt chat history, uploads and model request content at rest, with retention controls. Model providers and connected services still receive what the task needs. Review Settings › Data sharing and any separate consent to attach diagnostic evidence to a feedback report. ### How much does it cost? Creating an account is free, and a new Sunport starts with trial credit. Usage draws from one prepaid balance in dollars. OpenOrange adds no inference fee to calls using your own provider keys; your provider still bills you. Owners and billing admins can buy credit by card where the app offers it. Review the amount and applicable tax at checkout. For enterprise pricing, contact contact@openorange.ai. ## Access and contact OpenOrange is for individuals, teams and companies, and large companies already use it. Create an account to start a free trial in a private Sunport of your own. Companies can email us for custom terms. Start free: https://app.openorange.ai/sign-up?ref=home&cta=markdown Sign in: https://app.openorange.ai/sign-in?ref=home&cta=markdown Contact: contact@openorange.ai ## OpenOrange product guides - [What is OpenOrange](https://openorange.ai/what-is-openorange/) - [AI infrastructure](https://openorange.ai/ai-infrastructure/) - [Sunport](https://openorange.ai/operator-layer/) - [AI activity](https://openorange.ai/ai-observability/) - [Agents and models](https://openorange.ai/runtime-adapters/) - [Integrations](https://openorange.ai/integrations/) - [Model API](https://openorange.ai/model-api/) - [Compute](https://openorange.ai/compute/) - [Security](https://openorange.ai/security/) - [Costs and usage](https://openorange.ai/model-usage-accounting/) - [Privacy and audit](https://openorange.ai/privacy-and-audit/) - [OpenClaw hosting](https://openorange.ai/openclaw-runtime/) - [OpenOrange Code](https://openorange.ai/openorange-code/) - [Hermes hosting](https://openorange.ai/hermes-agent-hosting/) - [Connect any agent](https://openorange.ai/connect-any-agent/) - [Telegram, Slack and Mattermost](https://openorange.ai/telegram-slack-mattermost-agents/) - [Pricing](https://openorange.ai/pricing/) ## Available languages - [English](https://openorange.ai/index.md) - [Deutsch](https://openorange.ai/de/index.md) - [Español](https://openorange.ai/es/index.md) - [Français](https://openorange.ai/fr/index.md) - [Italiano](https://openorange.ai/it/index.md) - [Português](https://openorange.ai/pt/index.md) - [Русский](https://openorange.ai/ru/index.md) - [Türkçe](https://openorange.ai/tr/index.md) - [Bahasa Indonesia](https://openorange.ai/id/index.md) - [Tiếng Việt](https://openorange.ai/vi/index.md) - [हिन्दी](https://openorange.ai/hi/index.md) - [ไทย](https://openorange.ai/th/index.md) - [日本語](https://openorange.ai/ja/index.md) - [한국어](https://openorange.ai/ko/index.md) - [简体中文](https://openorange.ai/zh/index.md)