Nasdaq puts agentic AI inside Calypso for the trade lifecycle
Nasdaq said Tuesday it launched an agentic AI operating environment inside Nasdaq Calypso so banks and other institutions can run and connect AI agents across capital market and treasury workflows under sandboxing and oversight, starting with a natural-language assistant.
Banks already want agents on post-trade work, but they will not paste trade and collateral books into a random chatbot. Nasdaq is putting a sandboxed MCP layer inside Calypso so agents can touch the system of record under the bank’s own perimeter.
On Tuesday, 29 September 2026, Nasdaq launched a framework for putting AI inside capital market and treasury operations. At the center of that framework is an agentic AI operating environment inside Nasdaq Calypso. Agentic, here, means software that can take a step in a workflow, not only answer a chat question. An operating environment is the place those agents run. The Nasdaq press page stamps the release Sep 29, 2026, 4:00 a.m. Eastern. The dateline is New York, and the page carries the release through GlobeNewswire. The Amazon Web Services US Press Center reprints the same text the same day and does not print an hour. Those lines are Nasdaq’s.
What Nasdaq says the environment is for. It is a contained, governed space where a financial institution can run, connect, and scale AI agents across the trade lifecycle. Contained means the work stays inside a boundary. Governed means the institution’s own rules still apply. The trade lifecycle is the path a trade takes from the order through risk, collateral, and the books. The page calls the launch a milestone in Calypso’s plan to be the orchestration layer that connects agents across capital market and treasury workflows. An orchestration layer is the switchboard that lines those agents up with the work. Those lines are Nasdaq’s.
Who already uses the platform, as the release describes it. Nasdaq calls Calypso an advanced capital markets and treasury platform. Banks, brokers, asset managers, central banks, and other financial institutions use it for front-to-back office trading, risk management, and collateral workflows. Front-to-back means from the trade itself through the records that follow. Those lines are Nasdaq’s. The release does not name a Calypso customer for this launch, and it does not print a price.
How a client connects. The environment lets clients use Nasdaq Calypso’s own agents and connect their own AI through an integrated layer built on the Model Context Protocol, which Nasdaq shortens to MCP. MCP is a shared plug so an agent can call another piece of software. Nasdaq calls it an emerging industry standard for linking AI agents to software. A client’s own AI, in that sentence, is the models and tools the institution already runs, not a chatbot Nasdaq picked for them. Those lines are Nasdaq’s.
The guardrails. The environment applies operational boundaries, live oversight, and strict sandboxing, with no external data retention. A sandbox is a closed box: the agent works on the data inside it, and that data is not kept outside. Nasdaq says the design keeps agents inside the institution’s perimeter and policies. A perimeter, here, is the institution’s own boundary. Clients retain full control of their data. Those lines are Nasdaq’s.
What the agents can read. They can operate on core Nasdaq Calypso data: the platform’s system-of-record trading, risk, and collateral records. A system of record is the book the institution treats as the official copy. The environment can also take in more enterprise, market, and reference data. Enterprise data is the firm’s own. Market data is prices and related figures. Reference data is the standing labels on an instrument, such as what it is and who issued it. Those lines are Nasdaq’s.
The first capability. Today’s launch lines up with the first agentic tool in the environment: a natural-language assistant that lets a user query the platform’s data, documentation, and other information. Natural language means a person can ask in ordinary words. Nasdaq says the point is to deepen how teams use the platform, raise productivity on data analysis and document review, and speed decisions. The subhead says the assistant is there to automate analysis of that system-of-record data. Those lines are Nasdaq’s. The release does not name a second tool that is already on.
What is still a plan. Nasdaq plans a suite of governed agentic workers for the friction of manual processes that still sit across capital market workflows. A worker, in that sentence, is an agent aimed at one of those chores. “Plans” is the release’s word. It does not give a day for that suite. Nasdaq also says it will partner with individual clients as they work through their own AI governance and control requirements to bring these capabilities online. That is a rollout with each client. It is not a statement that every Calypso user is switched on today. Those lines are Nasdaq’s.
Where it runs. The agents will be hosted by Nasdaq and delivered through Nasdaq Calypso’s cloud environment, built on Amazon Bedrock. Bedrock is Amazon’s service for running AI models, with the controls a regulated firm expects. On-premises clients will also be able to connect to those same agents. On-premises means the software sits on the institution’s own machines. “Will be hosted” and “will also be able to connect” are the release’s wording. Those lines are Nasdaq’s.
What Nasdaq says this sits on. The launch builds on other AI inside Nasdaq’s financial technology platforms. Nasdaq Verafin’s agentic AI workforce is used by more than 800 clients. Nasdaq AxiomSL has AI-powered regulatory reporting. More than 800 is the Verafin figure. It is not a count of firms running the new Calypso environment. Those lines are Nasdaq’s.
The survey Nasdaq cites in the same release. A recent Nasdaq and Acuiti survey found that almost half of respondents pointed to challenges from manual intervention in their post-trade operations. Manual intervention means a person still has to step in. Fifty-six percent said they risk losing ground to competitors if they do not put AI or machine learning into their clearing operations. Fifty-six percent is a bit more than one respondent in every two. Clearing is the step that makes sure both sides of a trade complete. Those figures are the survey’s, as Nasdaq prints them. The release does not say how many people were asked.
Magnus Haglind, head of capital markets technology at Nasdaq, said institutions need trusted, governed infrastructure inside the capital market platforms they already run if they are going to adopt AI agents at scale: a place to connect agents from many sources to real workflows, data, and controls. He said that by integrating agentic capabilities and serving as the orchestration layer, Nasdaq is evolving Calypso from a system of record into an intelligent platform that automates activity at every stage of the trade lifecycle. That quotation is his, in the release. The named capability at launch is the natural-language assistant. The wider suite of workers is the plan above.
In plain terms, Nasdaq told clients on Tuesday that Calypso now has a sandboxed place to run AI agents on the official trading, risk, and collateral records, and to plug in a bank’s own agents through MCP. The first tool is a natural-language assistant for the platform’s data and documents. Nasdaq hosts the agents on Amazon Bedrock, and a firm that keeps Calypso on its own machines can connect. The release does not print a price or a Calypso customer name.
The picture is an olive and gold desk graphic of that stack. A natural-language assistant sits at the top. Under it is the MCP layer, then the system-of-record trading, risk, and collateral records, then Amazon Bedrock. It is a diagram of the launch. It is not a photograph of a trading floor.