Users can discuss investment ideas with the AI, research assets and give it instructions that involve putting funds to work. Unlike an assistant that stops after producing research or a trade idea, Virtuals said its system can continue monitoring information and act when predefined conditions are met.
One example shared by the project instructed an agent to open a $1,000 leveraged position whenever Ether moved 2% within five minutes and close it following a specified social media post. Virtuals presented the example as a demonstration of how natural language instructions can be converted into ongoing trading rules.
Virtuals App lets AI execute trading instructions
Trading automation is one of the main functions being tested in the app. Users can give the personal AI a task, define how much money it can use and specify its permissions before leaving the agent to carry out the instructions.
Virtuals said early testers are using the app to continuously follow markets and conversations that would otherwise require manual monitoring. Some have configured agents to execute trades in response to social media posts, while others are building dashboards that follow token sniper wallets from entry through exit.
Research, monitoring and execution take place within a chat interface. Group chats are supported as well, allowing friends to discuss an investment while the personal AI carries out research or actions from the same conversation.
Virtuals described the product as the result of several earlier experiments around letting AI manage money, including Agent Commerce Protocol, EconomyOS, Butler on X and agentic trading competitions.
The company said it had spent years testing different approaches before moving toward a personal AI equipped with both a wallet and market access.
Access to the iOS beta is currently available through invitations from existing users. Virtuals has not announced when the app will move into an open beta or become generally available.
AI agents are moving closer to direct control of funds
Virtuals App enters a market where crypto companies have been giving AI systems progressively more access to wallets, exchanges and trading tools.
In August, Binance introduced Agent OS and its MCP server, allowing compatible AI applications to use market data and perform authorized trading functions across spot, margin, Convert and futures products. Agents cannot withdraw assets externally, while account actions depend on permissions granted by users.
Revolut took a more approval based approach when it connected Revolut X to third party AI assistants in July. Supported assistants can retrieve market data, inspect balances, examine positions and prepare trades, but users must approve each transaction before execution.
MetaMask moved into the same area in June with an early access Agent Wallet that allows AI agents to carry out swaps, perpetual futures trades and other onchain transactions under controls set by the wallet owner.
Virtuals is testing a setup where the agent can continue acting after the user establishes the operating rules. The company has not provided detailed public information on the custody model, transaction approval process or safeguards that apply when an agent is allowed to trade automatically.
Virtuals has been building infrastructure for autonomous AI agents
Virtuals Protocol has spent much of its development around infrastructure that allows autonomous agents to transact and coordinate economically.
In March, crypto.news previously reported that Virtuals was integrating Agent Commerce Protocol with Arbitrum. The integration was designed to let AI agents transact natively on the layer 2 network and use decentralized finance applications.
Agent Commerce Protocol provides a framework through which agents can discover services, coordinate tasks and settle transactions. Virtuals App brings a related concept directly to individual users by giving a personal agent access to capital and instructions from a conversational interface.
Similar infrastructure has emerged elsewhere in crypto. Base introduced new wallet capabilities in May that let connected AI agents track portfolios, swap tokens, send funds and interact with supported applications. Base requires a separate user confirmation when an agent proposes a transaction, and expected asset changes are displayed before approval.
Virtuals has not said whether its closed beta uses comparable confirmation requirements for every action or whether agents can execute within a previously approved spending range without further authorization.
Virtuals App tests AI trading inside group chats
Virtuals is positioning the group chat as more than a place to request information from an AI assistant.
One demonstration shared by the project showed an AI monitoring a trading condition after receiving an instruction from a user. Another showed an agent being asked to build a dashboard tracking sniper wallets holding a particular token, including opening a browser, finding relevant wallets and monitoring their balances.
A separate example placed the personal AI inside a group conversation where participants were discussing investment information. Users could ask the agent to research an asset without leaving the conversation.
Virtuals said traders in the closed beta are already using the product to follow market activity and social media conversations throughout the day, automate trades around posts and track token wallets through their full trading cycle.
For now, participation remains restricted to invited iOS users, with existing testers able to provide invitation links to others seeking access. Pasted markdown
