Deriv releases its internal AI workflow platform, separating staff permissions from agent access

Broker Deriv has published Deriv One, the internal artificial-intelligence workspace it developed to make reusable agent skills available across departments. Announced in Dubai on 29 October and published in the company newsroom on 30 October 2026, the software is offered under the MIT open-source licence and is designed to run on an organisation’s own infrastructure. That makes this an infrastructure release, not an announcement that client trading accounts can now be operated by the software.[1]

Deriv says the project began when employees were building independent agent skills without a shared answer to who could use them. The solution is a central skills catalogue connected to role-based access, identity management, a secrets store, sandboxed execution and usage-cost tracking. Staff work from a workspace or a named agent; agents may have their own Slack applications and accountable owners. These are material design controls because allowing an employee to call a model does not automatically mean the model should be allowed to call every internal system.[1]

What the public release contains

The company reports that 54 employees contributed 221 skills. Those figures describe internal contributions, not a count of independently audited or production-certified automations. Its wider business target is to automate 75% of manual workflows by the end of 2026; that is a management target, not a result already achieved. Open-source availability likewise does not transfer Deriv’s organisational permissions or integrations to another firm.[1]

For brokers evaluating the platform, the main due-diligence points are who approves a skill, which client or business records it can access, whether execution is isolated, where model requests are routed, and how a failed action is reconstructed. The licence reduces restrictions on using the code, but it does not resolve financial-sector privacy, retention, outsourcing or operational-resilience obligations. Firms need to evaluate their own environment before allowing agents to touch payments, onboarding, dealing or compliance processes.[1]

The significance is that a retail brokerage has chosen to publish an internal software layer rather than simply announcing an AI chatbot. It offers a concrete example of the permissions and audit problem that arises when many business teams build model-driven tools at once. Deriv has not supplied an independent benchmark of productivity or a controlled comparison with competing enterprise-agent platforms, so neither can be inferred from the launch.[1]

Report an error or submit an update