Skip to content
Vanta — Subnets — AI — News

Vanta Introduces Glitch AI Agents to Build, Test, and Run Trading Strategies

The free beta connects plain-English strategy development with Vanta’s Bittensor-based trading network.

Table of Contents

Vanta has introduced Glitch, an AI trading copilot that turns a trader's idea into coded strategies, tests them against historical market data, and runs the version the trader chooses to deploy. The product is opening through a free, invitation-based beta, with access rolling out in waves.

Glitch is a sister brand to Vanta, the proprietary trading operation connected to Bittensor's Subnet 8, or SN8. Its initial trading integration will let users build strategies for their Vanta Trading accounts, combining strategy-development software with access to the network's trading evaluations.

The product brings several jobs into one workflow: translating a setup into precise rules, writing the code, testing alternative versions, reviewing weaknesses, and monitoring an automated strategy.

Separate Agents Build and Challenge Each Strategy

Traders can start with a plain-English market idea, existing indicators and alerts, TradingView Pine Script, or a running strategy. Glitch's multi-agent workflow divides development among specialized roles rather than assigning the entire process to one chat agent.

An architect creates variations on the idea, a developer codes them, and risk and execution agents check position sizing and tradability. Glitch backtests each version with fees and slippage, the difference between an expected execution price and the price actually obtained.

For example, Glitch illustrates how a request to buy a breakout can become three strategies using daily, four-hour, or weekly highs. Those variations turn a broad trading intuition into explicit conditions that can be inspected and compared.

A separate "Skeptic" agent reviews the results and retests strategies on unseen data. That out-of-sample check evaluates a version against market history that was not used to develop it. The review also looks for overfitting, where rules capture historical noise rather than a repeatable pattern, and look-ahead bias, where a test uses information that would not yet have been available during the trade.

Glitch's example flags a strategy that uses a four-hour candle's closing price before the candle has finished. It also checks whether transaction costs or a change in market conditions would undermine the strategy. These checks are built into the development process, with reasons attached to the agents' recommendations.

The trader can advance a version, request revisions, or reject it. Nothing runs until the user deploys it; after deployment, Glitch executes the selected rules automatically and displays the strategy's state, positions, and recent signals in one view.

A separate conversational agent, Lead Quant, is available on every plan. It converts a described setup into a live alert and checks how often that condition would have occurred historically, including warnings when the sample is small. Alerts can arrive in-app, by email, or through Telegram with a chart snapshot. Operator and Builder users can also import Pine Script, with an equivalence check against the original logic.

How Glitch Connects to Vanta and SN8

Arrash, who builds at Vanta and Glitch, described the relationship as Vanta providing the funding layer and Glitch providing infrastructure to improve strategy performance.

He said every Builder subscription includes a $1 million Pro account evaluation on Vanta Network and that Glitch revenue goes back to the network. The bundle ties the software subscription to a trading assessment on Vanta's platform.

Users can route strategies they develop in Glitch to their Vanta Trading accounts. The connection also gives SN8 a subscription-software business tied to participation in the trading network, alongside the work of developing and evaluating trading strategies.

Vanta plans to expand integrations to other brokerages and proprietary trading firms. The company expects to share more details in the coming weeks.

Free Beta, With Four Plans at Launch

The Glitch beta includes access to the agent swarm for building strategies. Full subscription plans and one-minute market data are scheduled for the wider launch.

Glitch lists the following launch pricing, billed month to month:

PlanMonthly PriceMain Capabilities
ExploreFreeLead Quant, historical alert checks, and up to three live alerts
Signals$19Up to 25 alerts, one-minute bars, and TradingView webhooks
Operator$39Automated strategies, Pine Script imports, and Lead Quant backtests
Builder$79Full agent-swarm strategy development, swarm backtests, and out-of-sample checks

Builder is also listed at $59 a month when billed annually, with a seven-day free trial. Paid plans begin at launch, and Glitch says launch prices may change.

Agent work uses credits, while live alerts and running strategies do not. Weekly allocations range from 50 credits on Explore to 450 on Builder; unused credits do not carry over, and paid subscribers can purchase additional packs. Each run shows a credit estimate before it begins.

At launch, traders will also be able to connect their own CryptoQuant, Coinglass, or Glassnode accounts. Glitch lists funding rates, onchain activity, order books, and price history among its data inputs, so strategy rules can draw on more than price indicators alone.

Comments

Latest