Circuit

Trade any signal.

Announcements at

[Circuit @CircuitAI]

what’s new at circuit?

$15.75B+ traded. new partners. faster agents.

describe it. paper trade it. fork it. go live.

put your ideas to work.

[Circuit @CircuitAI]

@chiragnarang23 🧠

[Chirag Narang @chiragnarang23]

I am nearly 5 months into the agentic trading space. A few thoughts on what I have learnt and what the future might look like.

Its becoming obvious that agentic trading will make the trading app a choice, rather than a requirement. You should be able to act on a thesis wherever you develop it: your favorite wallet or protocol, a chat assistant, a research terminal, or a DIY trading agent. Sometimes that means a single trade. Sometimes it means running a strategy for weeks, across several markets, without managing every action yourself.

Coding harnesses are an inspiration for how we get there. I want the same feeling from a trading harness: describe an idea, work through the missing decisions, test it and build something that runs. But once my strategy is built and funded, I want determinism at runtime, with explicit rules governing execution, permissions, and recovery. If the strategy needs model judgment, I should decide where that judgment applies and what it can change. Only after we make that reliable for everyday traders, we can gradually give frontier intelligence more discretion at runtime.

first wave of agentic trading started with MCPs but that alone won’t cut it, imo. Connecting an agent to a wallet and some venues gives it tools. We still need a system that can operate a funded strategy from start to stop. What happens if one part executes and another fails? When should it retry and when should it unwind? How do I backtest and paper trade it? I want those decisions worked through before my money moves. Users may not be comfortable leaving them to whatever the model decides at runtime. You can put that logic behind an MCP tool, but then you’ve built the trading system we’re talking about here.

Building a trading harness well requires getting five things right -

1/ World-class agents : Make it dead easy to turn a trading idea into a runnable strategy.

That could be as simple as a DCA, copying a wallet with your own position limits, or rotating capital between lending markets. Or as complicated as cross-venue funding arbitrage, delta-hedged stock LP, or systematic equity mean reversion. It could combine price, liquidity, news, X threads, prediction bets, onchain events as trading signals in ways we would never think of.

The platform needs to give agents the data and tools each strategy requires. That means providers such as DefiLlama and CoinGecko, direct venue data, block explorers, and specialized feeds for wallet tracking and copy trading, news webhooks... It needs to
choose sources based on the decision being made, fetch live data when needed, or inbuilt tools to minimize latency and better context management.

The agent also needs to understand the breadth of markets it operates in. We need to account for these differences across different kinds of dexes, lending protocols, perps, prediction markets, hybrid on/off chain protocols and cexs. Your agent should be able to not only buy/sell an asset, but LP, lend, hedge, find arbitrage, or any possible action that you can imagine.

The platform needs to build that knowledge into strategy creation and validation (backtesting, paper trade simulations) to test whether the agent implements what the user approved.

2/ It has to be snappy.

We are talking about changing a decades-old form factor from buttons and clicks to natural language prompting and dictation. But a prompt is not automatically a faster interface for every trade.

That sets a practical bar for speed. The platform needs to reduce the time spent interpreting requests, preparing trades, submitting them, and showing confirmed results. I expect more trading workflows, even discretionary one-time trades, to start with language. But we’ll have to earn that change in habit. The experience has to be meaningfully better than what people already use.

3/ Transaction lifecycle management, from the first allocation through the final exit and everything in between

Say I want to buy a tokenized stock on a dex, use it as collateral on Aave, borrow against it, and bridge the borrowed capital to Hyperliquid. There, I want to open a long HYPE position with tp/sl orders.

The agent needs to manage that entire lifecycle across three venues. It has to track collateral and debt on Aave, margin and liquidation risk on Hyperliquid, and the capital moving between them. It needs to monitor the exit orders and act within predefined liquidation guardrails.

When the perps position closes, the job is still unfinished. The agent needs to bridge the remaining capital back, repay the loan, and release the collateral. If the returned capital cannot cover the debt, the strategy needs an approved plan for covering the shortfall. Every step has to account for what actually executed before starting the next.

For trading agents, managing that entire lifecycle is the core job -

4/ best-in-class execution quality

Permissionless markets are a large part of what makes crypto interesting. But liquidity fragmented across issuers, protocols, and chains means more research for traders trying to get a good fill (and opportunities for arb strategies)

Sourcing best-in-class quotes will rely on existing aggregator infrastructure alongside direct venue integrations. Users shouldn’t need to research liquidity depth, funding rates, borrowing rates, slippage, and such. Agents need to translate traders’ intent into the best execution for their chosen asset, across available venues.

5/ TRUST

Users need control over what an agent is authorized to do without giving it unrestricted access to a wallet. Users will sit across a spectrum. Some will want self-custodial wallets with direct control. Others will trade on CEXs through API keys with delegated access. Either way, users need control over what an agent is authorized to do: what it can trade, which venues it can use, its budget, and when its authority ends.

Those permissions need enforcement at the wallet or exchange access layer. They cannot depend on the model remembering an instruction. The agent should have only the access needed to run the approved strategy, and users should be able to inspect and revoke that access.

This is what we're building at Circuit: a trading harness people can use directly and other teams can offer through their own products. They should be able to build for their users without rebuilding the entire trading system.

Becoming a standard for agentic trading is something we have to earn. if you're interested in building with us, DMs are open or try out here https://t.co/JHxqF5OQeh

[Circuit @CircuitAI]

@chiefkeith1000 @Enter_Elysium km chief

[Circuit @CircuitAI]

today, agents @Enter_Elysium

describe a strategy on circuit and an agent runs it on elysium around the clock.

memecoin sniper, dca, hypercore carry trade. you bring the idea, the agent does the work.

agents are among us.

[Circuit @CircuitAI]

circuit users are already building agents that copy the top traders on fomo, facilitate carry trades on stock tokens between Elysium and Hypercore, and DCA into $KNTQ through the launch window.

what will you build?

[Circuit @CircuitAI]

@Tekalign3773 @metricxyz looking into this

[Circuit @CircuitAI]

Circuit agents are now live on Robinhood Chain. Our first agent helps you provide liquidity for Stock Tokens paired with USDG on @metricxyz.

Each day, it ranks the pools by fees paid to liquidity providers and funds the top three. It checks them every minute, recenters when price moves, and steps aside before the market opens and around earnings.

let circuit carry the trade.

[Robinhood Crypto @RobinhoodCrypto]

When markets stop moving, Circuit agents keep going.

A new integration for Stock Tokens: now live on Robinhood Chain.

[Circuit @CircuitAI]

@metricxyz 🤝 we've enjoyed the partnership

[Metric @metricxyz]

Welcome to Robinhood Chain, @circuitAI!

Their first agent is live helping you to provide liquidity for Stock Tokens on Metric.

More liquidity means better prices for everyone trading them.

[Circuit @CircuitAI]

See how it works:
https://t.co/C2W3NQnGoz

[Circuit @CircuitAI]

@benjii @DeriveXYZ

[Circuit @CircuitAI]

Start with a prebuilt xStocks agent or fork one and make it yours: https://t.co/dWwBt0uLXZ

Not available in the US or to US persons. Geo restrictions apply. Read Kraken's xStocks Risk Disclosure at https://t.co/9BTUlCstfC.

[Circuit @CircuitAI]

The stock market just closed for the weekend. xStocks didn't.

Circuit agents now trade @xStocksFi on Solana and @krakenfx, nights and weekends included.

Describe a strategy in one sentence and an agent runs it.

[Circuit @CircuitAI]

@CoffeeNDrop @Enter_Elysium what are you talking about

[Circuit @CircuitAI]

agents are coming to @Enter_Elysium.

day one: describe a strategy on circuit and an agent runs it around the clock.

[Kinetiq @Kinetiq_xyz]

Autonomous agents from @circuitAI are entering Elysium.

e/acc

[foobar/ @0xfoobar]

Circuit will be the easiest way to trade on Elysium!

Prebuild your agents now to get ready for Monday 🟢

[Circuit @CircuitAI]

@0xfluflu feelings mutual 🤝

[Circuit @CircuitAI]

@DeriveXYZ 🤝 "the default way to trade"