Diagrid catalyst 2.0 raises bar for AI agent security

Diagrid Catalyst 2.0 significantly advances enterprise AI agent orchestration by introducing durable execution across more than 10 diverse AI agent frameworks, a critical feature that ensures workflows can pause and resume seamlessly over hours or even days without losing context. This capability supports complex, real-world processes involving loops, retries, conditional branching, and human-in-the-loop approvals, transcending simplistic linear sequences to meet the nuanced demands of enterprise operations.

Recognizing the heterogeneous nature of enterprise environments, Catalyst 2.0 offers robust multi-framework support that enables orchestration across in-house systems, SaaS platforms, and various cloud providers. This flexibility is essential as organizations rarely operate within a single, homogenous agent ecosystem, requiring a unified orchestration layer that coordinates diverse agents without imposing restrictive infrastructure assumptions.

Catalyst 2.0 elevates governance and observability by integrating data sharing, decision handoffs, and guardrails into a cohesive orchestration discipline rather than disparate tools patched together. Beyond the foundational capabilities of agent frameworks, it delivers enterprise-grade runtime, monitoring, access control, and audit trails, addressing critical gaps that previously hindered the transition from prototype to production—especially in regulated sectors demanding stringent compliance.

To future-proof interoperability, Catalyst 2.0 embraces emerging standards like the Model Context Protocol (MCP) and Agent-to-Agent (A2A), facilitating seamless communication among agents, tools, and platforms from different vendors without the overhead of custom integrations. This strategic alignment positions Catalyst as a linchpin in a growing ecosystem where cross-vendor collaboration is paramount for scalable, flexible AI agent orchestration.

Uniphar’s transition from the open source Dapr Workflows to Diagrid Catalyst 2.0’s commercial solution was driven primarily by the need for enhanced observability and governance in their AI agent orchestration. Catalyst’s integrated observability dashboards and centralized portal have significantly simplified Uniphar’s operational oversight, enabling real-time system performance monitoring and detailed error investigation, as noted by Oisin Vaclav Haken who emphasized the ease of developing and hosting agents with full visibility.

Catalyst 2.0’s commercial features provide Uniphar with critical capabilities for enterprise-grade governance, security, and regulatory compliance. These include cryptographic identities for authentication and authorization, declarative access policy enforcement, and auditable attestation that links AI agent actions directly to human managers or project groups. CEO Mark Fussell highlighted how this digitally signed verification supports compliance with emerging regulations such as the EU AI Act, ensuring operational transparency and accountability.

Uniphar is actively leveraging Catalyst 2.0’s flexible architecture to experiment with multiple AI agent orchestration frameworks, including Microsoft Agent Framework, while maintaining control over their own infrastructure. The platform’s agentic governance features allow Uniphar to easily activate controls that validate and constrain agent-to-agent communication, providing essential guardrails as their AI orchestration architecture matures. Haken underscored the importance of these governance capabilities in managing complex agent interactions securely and compliantly.

The rapid proliferation of AI agents as non-human identities within enterprises has created an urgent need for rigorous identity governance and policy enforcement to prevent over-privileging and the emergence of shadow AI identities. With AI agents often granted API keys and database credentials far exceeding what a human contractor would receive, organizations face heightened risks of unauthorized access and cascading breaches, especially as agents increasingly communicate autonomously without human oversight. As highlighted in the 2026 analysis 'The shadow identity crisis,' extending Zero Trust principles to these non-human identities is essential to continuously verify and govern AI agents, mitigating risks from implicit trust among interconnected agents.

Diagrid Catalyst 2.0 addresses these governance challenges by implementing a zero-trust framework that combines identity verification, policy enforcement, and observability to ensure safe, compliant AI agent orchestration. This IPOE model—Identity, Policy, Observability, and Evidence—enables operators to rapidly move AI agents into governed production with pre-runtime checks, runtime policy controls, and mediated connectivity trust, effectively mitigating threats such as impostor agents, prompt injection, and lateral movement. Telefónica’s Patricia Díez Muñoz encapsulates this approach: 'AI agents will only be trusted in telecom operations if operators can verify who is acting, control what is allowed, and prove what actually happened,' underscoring Catalyst’s role in bridging the trust gap in enterprise AI deployments.

Traditional security models relying on static credentials and broad platform permissions have proven inadequate for AI agents, which combine machine speed with human-like unpredictability, leading to a 4.5x higher incident rate when least privilege principles are not enforced. The Catalyst’s governance framework innovates by minting Subscriber Context Tokens that provide agents with just-in-time, scoped access to structured business context rather than direct database credentials, thereby reducing the blast radius of potential breaches. This proactive containment strategy, coupled with immutable audit trails and human-in-the-loop confirmation gates, not only enhances operational safety but also supports regulatory compliance in high-stakes environments like telecom.

The growing dominance of AI agents and other non-human identities—now outnumbering human users by over 80 to 1 in many organizations—has outpaced traditional IT governance, creating a widening AI confidence gap. Enterprises must unify human and machine identities under centralized zero-trust controls with just-in-time privileges and comprehensive lifecycle management to close visibility gaps and prevent unauthorized actions. As noted by identity security leaders like Okta and industry experts, identity governance is becoming the foundational security layer across all technologies, with budgets increasingly reflecting its critical role in managing the evolving threat landscape posed by autonomous AI agents.