1) Create your account
Sign in on ara.so, then open the console. Ara can be used from web and desktop.2) Add a model provider key
Ara supports BYOK for major providers:| Provider | Console |
|---|---|
| Anthropic (Claude) | console.anthropic.com |
| OpenAI (GPT) | platform.openai.com |
| Google (Gemini) | aistudio.google.com |
| OpenRouter | openrouter.ai |
3) Spawn your cloud session
Ara creates an isolated sandbox (cloud_session) for your user:
- Session is started through the cloud API (
/session/start) - Runtime boots agent workers + channel bridges
- Filesystem mounts under
/root/.ara/workspace - Realtime routes become available for chat and terminal
4) Connect one communication channel
Pick a channel and validate end-to-end flow:- WhatsApp: QR pairing
- Telegram: bot token
- Email: IMAP/SMTP credentials
- iMessage:
- Mac desktop bridge for local iMessage
- Linq bridge for always-on iMessage/RCS/SMS
5) Run your first autonomous workflow
Try a prompt that forces planning + tool use:“Create a skill that summarizes my unread email every morning, checks my calendar, and sends me a concise message.”This exercises:
- Skill creation and persistence
- Message handling
- Filesystem writes
- Multi-tool execution
6) Verify memory and recursion
Follow up with:“What did you learn from earlier tasks, and how would you improve this skill next run?”You should see the agent use persisted memory and refine behavior.
See the full architecture
Understand the app layer, cloud control plane, and sandbox runtime.