Nasdaq Calypso Builds Agentic Layer for the Shift to Always-On Markets

Nasdaq is extending its Calypso capital markets solution beyond its traditional role as a system of record, adding an agentic AI layer designed to support increasingly automated workflows across trading, risk, clearing and collateral management while retaining the deterministic processing at the core of the platform.

For Magnus Haglind, Head of Capital Markets Technology at Nasdaq, the distinction between the existing processing environment and the new intelligence layer is fundamental.

“We’re moving from being a trusted system of record – and we’re going to continue to be that – to becoming an intelligent platform of action,” he tells TradingTech Insight. “We have proven programmatic, deterministic, rules-based workflows, and that’s what people want to keep. But there’s also a lot of human intervention and decision-making during processing. What we’re introducing is an agentic layer around those workflows to help the operator become supercharged and informed, helping with more efficient and intelligent decision making.”

Extending the intelligence layer

Rather than replacing Calypso’s established workflow engine, Nasdaq is extending the platform horizontally through a dedicated analytics and data-management layer. This can hold time-series information, intermediate calculation results that might not previously have been retained, technical and operational metrics and clients’ own data.

That broader data layer also expands the boundaries within which agents can operate. MCP provides a standard interface between agents and individual systems, while an A2A protocol is intended to support communication and orchestration between agents as multi-agent workflows develop.

The initial applications are deliberately more constrained. One example cited by Haglind is an operations user logging into Calypso and being presented automatically with the trade breaches carrying the greatest risk exposure, rather than having to search through exceptions and determine manually where to start.

Building towards bounded autonomy

Nasdaq’s architecture is designed to support agents capable of taking actions, but Haglind says the transition towards autonomy will be gradual.

“We’re building a framework that can be fully autonomous, but we’re not starting with unleashing these capabilities without guardrails.,” he says. “The next step will be to increase the degree of autonomy across the agents, but initially we want to introduce this in a way that’s consistent with regulatory and risk-based principles.”

Exception management and reconciliation are among the likely early candidates for greater autonomy because firms can define mandates around the economic consequences of an agent’s actions. An institution could determine which exceptions an agent is permitted to resolve automatically and where human intervention remains necessary.

Agents also operate under Calypso’s existing access controls. They cannot access a market, account or object for which they haven’t been permissioned, allowing firms to extend existing control structures rather than creating a separate privileged route into the platform.

Telemetry will provide another layer of control, allowing actions to be reconstructed and audited, including the data and reasoning involved. Haglind expects banks ultimately to feed this information into enterprise-wide monitoring environments as agents proliferate across different systems.

The always-on challenge

The move towards extended-hours and potentially continuous markets adds another dimension. HSBC recently expanded its use of Calypso from OTC derivatives and repo clearing to exchange-traded derivatives and plans to extend it to US Treasury cash transactions. The bank cited real-time visibility across clearing activity, margin, collateral and risk as it prepares for developments including always-on markets and tokenised assets.

For Haglind, continuously available trading creates a considerably broader operational problem than simply keeping execution systems running for longer. “You’ll never be able to staff yourself out of an always-on situation. You’re going to need to utilise this technology. For most people, it’s relatively easy to create a vision of what always-on markets mean – basically, I can trade around the clock. Always-on risk management is much more complicated.”

The challenge extends across liquidity, treasury, currencies and collateral, all of which have traditionally been managed around established market sessions and operational cycles. As those boundaries become less distinct, firms need to understand their exposures and available resources continuously rather than around an end-of-day snapshot, notes Haglind.

“Do I have liquidity? Do I have enough collateral at the right point at the right time? Do I have the right liquidity in US dollars, other currencies or Treasuries? Do I understand my risk exposure? It’s going to require much more informed decision-making. Always-on markets will lead to always-on risk management,” he says.

This creates a potentially important role for agentic technology beyond productivity improvements. As trading hours expand and operational windows contract, the ability to identify, prioritise and eventually resolve bounded problems autonomously could become part of the infrastructure required to support continuous markets.

While Calypso’s existing processing workflows will remain deterministic, agents may now take on more of the investigation, prioritisation and decision-making around them. For Nasdaq and its clients, the next step is establishing where agents can safely move from identifying and recommending an action to carrying it out.