The new capabilities enable financial institutions to design, govern, and deploy compliant AI agents within their Performativ environment or existing enterprise infrastructure.

Copenhagen, Denmark, November 5, 2025: Performativ today announced the launch of Custom Agents and the AgentKit Builder, two new capabilities designed to help financial institutions integrate AI safely and compliantly into their core operations.

Built on the OpenAI AgentKit foundation and powered by Performativ’s Model Context Protocol (MCP) architecture, the solution allows banks, capital markets firms, and asset managers to design, govern, and deploy AI agents within their existing workflows while maintaining full regulatory oversight, auditability, and control.

“AI must operate under the same governance frameworks as the rest of the enterprise,” said Albert Geisler Fox, CEO of Performativ. “With Custom Agents, institutions can now deploy intelligent automation that is both production-ready and compliant by design. Every action is permissioned, explainable, and accountable.”

‍

Compliance Engineered by Design

Performativ’s Custom Agents introduce a structured framework for enterprise AI governance. Each agent operates through an MCP user with least-privilege, role-based access controls, ensuring that data access, processing, and actions remain within approved boundaries.

All agent operations, including inputs, tool calls, parameters, and outputs, are logged and versioned, producing a complete chain of evidence suitable for regulatory review.

Administrators can define and enforce policy guardrails that determine when an agent can run, what data it can read or write, and under which approvals. Structured output schemas ensure predictability and traceability, allowing institutions to deploy AI-driven workflows that meet internal model governance standards and align with the European AI Act’s requirements for human oversight and explainability.

‍

Enterprise Deployment Across Complex Estates

Custom Agents are designed for large-scale deployment in institutions with distributed architectures, multiple systems of record, and stringent compliance mandates.

Agents can operate as native Performativ activities such as tasks, alerts, or scheduled operations or as custom actions that integrate directly with third-party or legacy systems under the same policy envelope.

For larger institutions managing complex data estates, Performativ’s CenterStage technology enables seamless integration without replacing existing infrastructure. It unifies custody, trading, and client relationship data, providing a governed execution layer for AI workflows that preserves data lineage and internal control frameworks.

“Enterprises need to move from experimentation to deployment without compromising governance,” said Geisler Fox. “Custom Agents make that transition possible by embedding compliance, security, and accountability into the runtime itself.”

‍

Operational and Regulatory Readiness

The platform’s auditability and policy enforcement are designed to support the supervisory expectations of financial regulators across Europe.

Performativ’s approach addresses key enterprise requirements, including:

• Traceability and evidence capture for every model-driven decision or recommendation.
• Segmentation by business unit or legal entity, ensuring jurisdictional and data-boundary compliance.
• Human-in-control architecture, where all agent actions can be reviewed, approved, or halted by designated officers.
• Policy alignment with AI governance frameworks, including internal model risk controls and the European AI Act.

These capabilities allow institutions to deploy AI responsibly, with full operational resilience and transparency.

‍

Demo Showcase

A recorded demonstration of Custom Agents and the AgentKit Builder illustrates how institutions can, for example, build a virtual analyst that classifies liquidity needs, checks policy guardrails, and generates compliant investment proposals all within a governed, auditable framework.

The demo can be watched below.

‍

Program Access and Availability

Custom Agents and the AgentKit Builder are now available in controlled preview for financial institutions, and Performativ will release additional documentation in the coming weeks.

Institutions interested in participating can contact Performativ here.