Getty Images

Uniphar taps Diagrid Catalyst for orchestration, observability

The company's platform engineering team replaced open source Dapr with the commercial product for built-in observability, with AI agents waiting in the wings.

After building a cloud cost management system using an open source workflow orchestration tool, a global healthcare services company moved to the commercial version with an eye toward future AI agent orchestration.

Uniphar Group, headquartered in Dublin, operates in more than 160 countries, with subsidiaries serving more than 200 partners in the pharmaceutical, biotech and medtech industries. Its platform engineering team at Dublin headquarters supports a Microsoft Azure cloud infrastructure and application stack written in C# and .NET. About two years ago, that team wanted to dig deeper into cloud cost management than Microsoft's built-in tooling supported. It turned to an open source distributed application orchestration framework called Dapr Workflows to coordinate a monthly cost management reporting process comprised of more than 400 individual workflows.

That system, called Mammon, enabled Uniphar's platform engineers to obtain detailed resource-usage and allocation information while tying it to internal budget initiatives for business users, said Oisin Vaclav Haken, a contractor working as a DevOps consultant at Uniphar who helped lead the project.

"We can go even beyond certain resource types, actually look at a system's internal metrics and use those metrics to further subdivide costs," Haken said in an interview with TechTarget. "A hierarchy of workflows is a really great way to implement that … and it's all [configured using] business logic. There's very little code for the orchestration description, and it just simplified the whole thing tremendously."

Enter Catalyst, observability and AI agents

Diagrid, founded in 2021 to offer enterprise support for Dapr tools, first rolled out Diagrid Catalyst in late 2023, adding SaaS-hosted APIs and built-in governance, security and observability features. Since then, Diagrid has added agentic AI features to Catalyst, including durable and verifiable workflow execution with auditable attestation and governance. Catalyst 2.0, released July 28, provides cryptographic identities for authentication and authorization of apps, agents and MCP servers, and supports declarative access policy enforcement for AI agents.

It was primarily Catalyst's integrated observability dashboards that prompted Uniphar to begin moving to the commercial product from open source Dapr Workflows earlier this year, Haken said. That process was just completed this month.

"Now we have our own portal on top of the source data, so it just makes it easier for us to get a glance at a system's performance," he said. "If there are any errors we need to be concerned with, we can dive into the problem, see the details and investigate."

Agentic AI orchestration and governance is also on Haken's mind as his team begins experimenting with Microsoft Agent Framework for developer onboarding and offboarding workflows. for developer onboarding and offboarding workflows.

"It's very simple for us to develop these agents, then host them through Catalyst, and we get that observability," he said. "We looked at Microsoft Foundry, but we have invested heavily in our own infrastructure to host our agents, and [Catalyst] allows us to do that."

AI agent orchestration is still at an early, experimental stage at Uniphar, but as it matures, Catalyst's agentic governance features could come into play, since they can be turned on with a simple configuration change in the existing system, Haken said.

"There's really a big emphasis on kind of controlling agent-to-agent [communication], validating, constraining and having guardrails in place," he said. "At this stage, we're still looking at the architecture, implementation and design principles. But obviously, it's something that is in the back of our heads."

Diagrid Catalyst 2.0 adds verification amid AI regulation

Initially, Catalyst supported the Dapr Agents framework, but version 2.0, rolled out July 28, broadened that support to 10 more popular AI agent frameworks, including LangGraph, Microsoft Agent Framework, Google Agent Development Kit, AWS Strands, OpenAI Agents SDK and CrewAI.

Catalyst previously supported durable execution for AI agents -- if a specific task failed, it could restart and ensure it finished. Version 2.0 will now provide digitally signed verification that execution has finished and link agentic actions to individual human managers or project groups within an organization, said Mark Fussell, CEO of Diagrid, in an interview with TechTarget.

"There's a lot of legislation coming in, like the EU AI Act, that are saying … anything that touches the personal data of an EU member citizen, such as a loan application, for example … you're required under their law to show direct proof that the loan origination was approved and happened, and that means that you need digital verification," Fussell said.

AI agent governance is a hot topic among enterprises as pilot projects struggle to reach production and new regulations proliferate, according to analysts.

"AI governance is moving from a policy discussion into an operational requirement, but adoption is still uneven," wrote Paul Nashawaty, principal application development analyst at Efficiently Connected, Inc. (ECI), based in Holden Beach, N.C., in an email to TechTarget.

A July 2026 ECI Research survey of 320 technology professionals found that 93% of organizations are adopting AI dev tools, yet only 12% have implemented standard security and operational governance controls for those workloads.

Diagrid, a privately funded startup, remains primarily associated with Dapr Workflow orchestration, and must compete with much larger, higher-profile AI governance tools for enterprise attention. But the vendor has an opportunity to broaden its appeal in an increasingly complex regulatory environment with Catalyst 2.0, Nashawaty said.

"Diagrid fits on both sides of the aisle: Developers get to write agents in whichever framework is trending, while platform engineering, SecOps and legal teams get the tamper-evident logging, resiliency and compliance infrastructure required to take those agents safely into production," he said. "Organizations need to know which agent is acting, what application it can reach, which MCP server it can access, and even which individual tools it is authorized to invoke. … Catalyst aligns well with where AppDev security is heading."

At the same time, Catalyst is still evolving. Haken said he's waiting for it to support managing AI agent skills alongside Microsoft Agent Framework agents. AI agent skills are a set of instructions that developers use to direct AI agents to perform specific tasks; skills can also be shared among organizations, stored and governed in their own skills repositories.

"Skills are, to us, a very interesting way to separate sub-processes and their handling by the agent; if the agent hosts a fairly wide set of functionality, it's that much easier to maintain and modify skills without affecting everything else in the process, so it's a good feature to have there," Haken said.

Diagrid officials said this week that the update is under development and will be available soon.

Beth Pariseau, senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism. Have a tip? Email her or connect on LinkedIn. 

Dig Deeper on Application Architecture