About avalanche
AvaKit is an open-source toolkit for building Avalanche dapps, aimed at developers and AI coding agents. Its packages combine app scaffolding, wallet and chain helpers, and MCP access for scaffolding, deployment, and chain-state queries.
Key features
- 01Scaffold dapps from eight templates
- 02Social-login wallet components
- 03Deploy contracts from the browser
- 04React components and chain helpers
- 05Local development dashboard
Use cases
- 01Start a social-login Avalanche dapp
- 02Create an NFT minting application
- 03Deploy a contract and mint on Fuji
- 04Inspect local devnets and chain data
MCP support
The @avakit/mcp server lets Claude Code and Cursor scaffold apps, deploy contracts, and read Avalanche chain state. @avakit/studio also includes an MCP server.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
6Saved tool list · updated 2w ago
This tool takes no parameters.
Details
From the project READMEAvaKit 🔺
The open-source, AI-native developer toolkit for Avalanche.
Scaffold a social-login dapp, deploy-ready, with agent context baked in. One core, four surfaces. No seed phrases, no boilerplate.
avakit.dev · Documentation · Templates
▶ Watch the 90-second demo — one command to a live, social-login Avalanche dapp: sign in with Google, deploy an NFT from the browser, and mint it on Fuji.
npm create avalanche-app@latest
That's it. Connect with a social login, read your balance, and send your first transaction on Avalanche in minutes.

Then deploy an NFT contract and mint it, straight from the browser, on Fuji:

Proven live on Fuji. The mint in that clip is a real on-chain transaction; view it on Snowtrace (contract
0x7612…7786). No Foundry, no backend: the bytecode is bundled and deployed from the user's wallet.
Why AvaKit
Avalanche's C-Chain is EVM-compatible and end-user onboarding is already solved (Core wallet's seedless social login). The remaining friction is on the developer side: spinning up a modern dapp with onboarding wired up still takes hours. AvaKit removes that.
- Zero-setup onboarding: a temporary in-browser wallet appears so a newcomer can try a real Fuji transaction with nothing installed, and social login (Google, no seed phrase) or Core/MetaMask is the upgrade — for those, keys stay in the provider's HSM and AvaKit never touches them. (The burner is the exception, and holds its key in
localStorage— see SECURITY.md.) - AI-native by default: every generated app ships
CLAUDE.md,llms.txt, and.cursor/rules, and@avakit/mcplets Claude Code / Cursor scaffold, deploy, and read chain state for you. - shadcn/ui, themed: a clean design system with dark/light from day one. Copy-in components, no vendor lock-in.
- Deploy-ready: contracts compile to bundled bytecode, so you can deploy straight from the browser. Fuji testnet by default.
- Safe defaults: testnet-first, mainnet is explicit opt-in, secrets stay in env.
- Wrap, don't rewrite: built on viem, Web3Auth, and Foundry, packaged for a great DX.
Packages
| Package | What it is |
|---|---|
@avakit/core | Framework-agnostic kernel: viem clients, wallet adapters, deploy helpers, chain data |
@avakit/react | <ConnectAvalanche> social-login widget, <TransactionButton>, and hooks, built on shadcn/ui |
create-avalanche-app | Batteries-included scaffolder (8 templates) |
@avakit/mcp | MCP server: scaffold, deploy, and read Avalanche from Claude Code / Cursor |
@avakit/studio | Local dev dashboard: spin up devnets, send Interchain messages, inspect data (npx @avakit/studio); also an MCP server |
Templates
npm create avalanche-app@latest my-app -- --template nft-mint
| Template | What you get |
|---|---|
minimal | Social-login wallet, balance, and a first transaction |
nft-mint | Deploy an ERC-721 from the browser, then mint |
token-gated-app | Unlock content for holders of an access-pass NFT |
erc20-token | Deploy an ERC-20, mint supply, and transfer |
icm-messenger | Send a message between two Avalanche L1s over Interchain Messaging, on a one-command local devnet |
eerc-token | Register, mint, and privately transfer tokens with hidden balances (Encrypted ERC) |
l1-launch | Launch your own Avalanche L1 with one command, then explore it in a built-in dashboard |
token-bridge | Bridge an ERC-20 between two Avalanche L1s with Interchain Token Transfer (ICTT) |
Every template ships with a social-login wallet, shadcn/ui (black & white, dark/light), and AI context files.
Use it in an existing app
npm install @avakit/react @avakit/core viem
"use client";
import { injectedAdapter } from "@avakit/core";
import { fuji } from "@avakit/core/chains";
import { AvaKitProvider, ConnectAvalanche } from "@avakit/react";
export function App() {
return (
<AvaKitProvider chains={[fuji]} adapters={[injectedAdapter()]}>
<ConnectAvalanche />
</AvaKitProvider>
);
}
Add social login by installing @web3auth/modal and passing web3authAdapter({ clientId }) (from @avakit/core/web3auth).
@avakit/coreand@avakit/reactare ESM-only (useimport, notrequire; Node ≥ 20.11).
Build with an AI agent
Add the MCP server to Claude Code, Cursor, or Claude Desktop:
{
"mcpServers": {
"avakit": { "command": "npx", "args": ["-y", "@avakit/mcp"] }
}
}
Then just ask: "Scaffold an nft-mint dapp and deploy it to Fuji."
Repository layout
packages/
core/ @avakit/core
react/ @avakit/react
mcp/ @avakit/mcp
studio/ @avakit/studio (local dashboard + MCP)
create-avalanche-app/ scaffolder + 8 templates
apps/web/ the avakit.dev site (EN + TR)
examples/ live demo dapps (hello-avax, web3auth-demo)
docs/ planning, PRD, architecture, ADRs, specs, security review
Contributing
pnpm install
pnpm build # all packages (Turborepo)
pnpm test # Vitest
pnpm lint # Biome
pnpm typecheck # TypeScript
Requirements: Node ≥ 20.11, pnpm ≥ 10. See CONTRIBUTING.md and the design docs in docs/. Conventions: English everywhere, shadcn/ui only, Framer Motion / GSAP for animation, latest stable versions.
License
MIT © AvaKit contributors
Built on viem, Web3Auth, Foundry, and shadcn/ui. Testnet-first (Fuji); mainnet is opt-in.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.