Getting started
What is Aomi?
Aomi is a hosted runtime that turns your APIs into AI-callable tools and ships them as an embeddable chat experience, with simulation-first onchain transactions and non-custodial wallets built in. See What is Aomi? for more.
Who is Aomi for?
Teams building agentic AI on top of blockchain APIs like DeFi protocols, exchanges, prediction markets, wallets, and analytics tools. Anyone with a REST API and a need for a chat interface that can read onchain state and propose transactions.
How do I get started?
Run the Quickstart to install the widget into a Next.js app in one command, or read the Integration Guide for the full pipeline.
Do I need a backend?
You need the APIs you want exposed as tools. The Aomi runtime is hosted, so you don’t run LLMs, the simulation pipeline, or session storage yourself.
Can I try Aomi without writing code?
Yes. Use the Web App or the Telegram bot with the default App, no API key required.
What is an Aomi App?
An App is an isolated agent environment: its tools, system prompt (preamble), model, and API key. Switching Apps changes the agent’s full capability surface. See Apps & Authentication.
A tool call is the AI invoking one of your registered APIs or on-chain actions. When the assistant determines it needs data or to perform an action, it calls the appropriate tool, processes the result, and responds naturally.
Yes. Tools can execute concurrently when the model issues parallel calls.
Use the Rust SDK to define a tool that wraps your API. See Custom Tools.
Can I use my existing OpenAPI / REST endpoints?
Yes, that’s the intended path. Wrap each endpoint as a tool. A one-to-one mapping isn’t required; you can group multiple endpoints into a single tool for cleaner LLM context.
Models
What AI models are available?
Claude (Anthropic), GPT-4o (OpenAI), and 100+ models via OpenRouter. Switch at runtime without redeploying.
How do I change models?
Use the model selector in the ControlBar (widget, web app) or the /model command (Telegram).
Can I use my own LLM API key?
The hosted runtime uses Aomi-managed LLM credentials. For self-hosted deployments or BYO-key arrangements, contact us.
Wallets & transactions
Is my wallet safe?
Yes. Aomi never holds private keys or funds. Every transaction is simulated on a forked network before it reaches your wallet. You review token changes, gas, and contract calls before signing.
Can I use Aomi without a wallet?
Yes. Reads, price checks, and most queries don’t need a wallet. Wallet connection is only required for on-chain transactions.
What wallets are supported?
Para handles connect and account management; signing flows through wagmi-compatible connectors (MetaMask, WalletConnect, Coinbase Wallet, and others).
What is simulation-first execution?
Before any transaction reaches your wallet, Aomi forks the current network state with Anvil and simulates the transaction. You see the exact outcome before approving. See Simulation Reference.
What chains are supported?
Ethereum, Base, Arbitrum, and any EVM chain you configure. See the Architecture page for setup.
Supported via Account Abstraction for smoother UX without custody changes.
Sessions & threads
How do threads work?
Threads organize conversations into separate sessions. Each maintains its own history, context, and wallet binding. Create, switch, rename, and delete them from the sidebar.
Why did the assistant refuse my request?
The assistant operates inside the active App’s constraints. If the preamble restricts certain actions, or the required tools aren’t available in the App, it refuses. Switch Apps if you need different capabilities.
How long are sessions retained?
Session and thread history is persisted per the runtime’s configuration. See Sessions Reference.
Integration & deployment
Widget: fastest path, full chat UI, edit copied source files. Headless: full control over UI, build with @aomi-labs/react hooks. See Frontend Setup for the comparison.
Yes. See the MetaMask example for one pattern.
Does Aomi work in non-React frontends?
The runtime API is plain HTTP + SSE, so any frontend can talk to it. The @aomi-labs/react hooks are React-only; for other frameworks, use the API directly.
Pricing & access
How much does Aomi cost?
Pricing varies by usage and App configuration. Contact us for a quote.
How do I get an API key?
Contact us or use the default App in the Web App and Telegram bot for testing.
Troubleshooting
How do I report a problem?
Reach out through the Telegram bot or aomi.dev/contact.
What data does the assistant have access to?
Only the tools and data sources configured in the active App. It cannot access data outside the App’s scope. Conversation history is retained per session for context.
Common causes: the tool isn’t registered in the active App, the description is unclear, or required parameters aren’t typed. See Troubleshooting.
Next steps
Last modified on June 4, 2026
