Two concepts are central to how Aomi works — worth understanding before you dive in.
Agentic Application
The isolated agent environment that defines what the AI can do and access. Each App has:
- Tools — the functions the agent can call (your APIs, on-chain actions, data sources)
- Preamble — the system prompt that shapes agent behavior and constraints
- Model — the LLM powering the agent (Claude, GPT-4o, or others)
- API key — scoped credentials, isolated per App
Both users and builders interact within an App. Switching Apps changes the agent’s tools, context, and behavior entirely. Running multiple Apps means running multiple independent agents.
Non-Custodial Automation
Aomi never holds your private keys or funds. Every transaction is simulated on a forked network before it reaches your wallet, so you can review the exact on-chain outcome before approving.
This simulation-first approach is the key safety mechanism. Users see exact token changes, gas costs, and contract calls before signing anything. The AI constructs and proposes transactions; the wallet signs them locally.
Integration Paths
Builders have three ways to integrate Aomi:
| Path | Best For |
|---|
| Widget (shadcn) | Quick integration, standard chat UI, projects using shadcn/ui |
| Headless Library | Fully custom designs, non-standard layouts, maximum control |
| Telegram Bot | Reaching users on Telegram without frontend deployment |
@aomi-labs/react package. The widget is built on top of the headless library.
For Users
Access Aomi through any supported channel:
- Web App at chat.aomi.dev — full chat interface
- Telegram — @aomi_sendit_bot
- Discord — coming soon
- iOS — coming soon
For Builders
Integrate Aomi into your product:
- Quickstart — get a working widget in minutes
- Integration Guide — full-stack walkthrough with examples
- API Reference — complete HTTP API documentation
- CLI — validate prompts, inspect sessions, simulate execution
Next Steps
Last modified on May 29, 2026
