Situation
Trading does not close. It is multi-venue, and a slow runtime, a leaked identity or a trace you cannot read shows up in days. That is why we built here first, on our own product, with real orders.
Ours · live markets
Warren
Warren is an autonomous fund manager built on Kodeus and run on live markets before we asked anyone else to trust the platform. Users set capital and risk limits. Warren researches, executes, monitors and manages the portfolio inside those limits, against live venues, around the clock. Quant strategists can publish strategies that users allocate to.
What it does
Two modes. In copilot, Warren recommends and the user approves each action. In autopilot, Warren executes inside the guardrails the user set. A proposal that breaks a limit does not become an order.
What the runtime did
Every trade passes the same guardrails, approvals and traces a customer product would use. In the published trace, figures redacted, the agent proposes a place_order. The position check passes. The daily loss limit holds. The order is not placed, the user is notified, and the verdicts, latency and cost are written.
What we publish
The use cases page carries the current figures: 100+ traders, $3.0M+ paper volume, $300K+ live volume, 55+ strategies, and 16+ exchanges and prediction markets. The homepage labels the $300K+ figure live capital. We do not publish a return.
Enterprise finance
Nash
Nash is long-running workers for finance operations. They monitor systems, reconcile records and escalate to a person when policy says they must. Same runtime as Warren. Different tools, different policy.
Situation
Finance operations work does not fit in a single chat turn. The worker has to stay up, notice a mismatch, and stop when the spec says a person has to decide.
What it does
It watches the systems it is connected to, reconciles records, and escalates instead of improvising past a rule. The escalation is the product, not a failure of the worker.
What the runtime did
Identity stays on the request. Policy can hold the action. The trace records the verdict. Nothing here depends on a shared service account or a prompt that asks the model to be careful.
What we publish
The description above is the whole public account. We do not add volume, headcount or a customer quote for Nash.
Research partnership
Embodied agents with ARTPARK at IISc Bangalore
A research partnership using the Kodeus runtime for embodied agent work. The same identity, policy and trace model, applied to agents that act in the physical world.
Situation
An agent that moves something in the world has the same need as an agent that moves money: you have to know who it acted for, what it was allowed to do, and what it actually did.
What it does
The partnership applies the runtime to embodied agents. The tools change. The operating layer does not.
What the runtime did
It supplied the identity, the policy and the trace. Those are the parts we claim. We do not describe lab results, papers or hardware that are not already on this site.
What we publish
That it is a research partnership with ARTPARK at IISc Bangalore, and that it uses the same model as the other runs on this page.
What is not a case study
The use cases page has industry cards: support, healthcare, insurance, operations, research. Those are not customers. They are the shapes we see most often when a team describes what it wants to ship. Each one breaks without per-user isolation, a policy that can refuse, or a record that holds up later. If your company is not named on this page, it is not a Kodeus case study.
Warren’s figures live on the use cases page, where they are counted in the page rather than restated as a new result.
Frequently asked questions
Is Warren a customer?
No. Warren is ours. It is an autonomous fund manager we built on the runtime and ran on live markets before asking anyone else to trust the platform.
What figures do you publish for Warren?
The use cases page publishes 100+ traders, $3.0M+ paper volume, $300K+ live volume, 55+ strategies, and 16+ exchanges and prediction markets. The homepage labels the $300K+ figure live capital. Those are the figures. We do not add a return number.
What is Nash?
Long-running workers for finance operations. They monitor systems, reconcile records and escalate to a person when policy says they must. Same runtime, different tools and policy.
What is the ARTPARK work?
A research partnership with ARTPARK at IISc Bangalore, using the Kodeus runtime for embodied agents. The same identity, policy and trace model, applied to agents that act in the physical world.
Are the industry cards on the use cases page customers?
No. They are the shapes teams describe when they say what they want to ship. This page only names Warren, Nash and the ARTPARK partnership.
Where is the Warren trace?
On this page and on the use cases page. The figures in the sample trace are redacted. The daily loss limit holds the order, the user is notified, and the verdicts are written.
Do you publish customer quotes?
Not on this site. A case study here is a description of work we can name, with only the facts already published.
How do I see a traced turn?
Book a demo. We will walk one turn, including a call the policy refused.
Walk a turn like Warren’s
Book a demo. We will show a held action and the trace that recorded it.