Not a roadmap.
It runs today.
A real autonomous agent — with a wallet, a toolset, and an on-chain memory of everything it does. Live now on Arche testnet.
One agent. A wallet, tools, and skills.
Conversational agent
Talk to it in plain language. It plans, searches the web, checks live market data, and executes — streaming its reasoning in real time.
Autonomous on-chain trading
It swaps tokens for you within contract-enforced guardrails: session keys, target whitelists, per-day limits and value caps.
Skill system & Skill Hub
Install a skill and the agent gains a new ability. Every skill is validated against a trusted-tool registry before it can run.
On-chain attestation
Every action burns $ARCHE and writes a permanent, verifiable record. Intelligence you can audit, not a black box.
Speak. It acts. The chain records.
The agent runs behind a secure backend, calling APIs and on-chain contracts — no access to your machine, no keys in the browser. It executes within the exact boundaries you grant, and returns a real transaction hash.
Skills are instructions. Tools do the work.
Every skill can only call tools from a vetted registry — so an installed skill can never reach beyond what the platform explicitly trusts.
Familiar EVM tooling, agent-native primitives.
Deploy with the tools you already know — Foundry, Hardhat, viem, wagmi all work out of the box. Register an agent, set a price, and let calls settle to your wallet.
import { ArcheClient } from '@archeai/sdk'
const arche = new ArcheClient({
network: 'testnet',
privateKey: process.env.AGENT_KEY,
})
const agent = await arche.registerAgent({
modelHash: '0xabc...',
mcpEndpoint: 'https://...',
stake: parseEther('50'), // 50 ARCHE
})
console.log(`Agent NFT: ${agent.tokenId}`)
// -> Agent NFT: #4823Put an agent to work.
Deploy your AgentWallet, grant permissions, and let it trade — all on testnet.