Diagrid Catalyst 2.0 adds durable execution to more than 10 agent frameworks
SiliconANGLE Duncan Riley ● Covered by 2 sources
Diagrid launched Catalyst 2.0, a plug-in that makes AI agents automatically recover from failures without rewriting code. It also cryptographically verifies every step, so businesses can actually trust and prove what agents did.
Based on reporting by SiliconANGLE, Duncan Riley — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Building an AI agent is one thing. Getting it to survive a crash without a human babysitting the process is another problem entirely, and it's the one Diagrid Inc. is now going after with Catalyst 2.0, released today. The update to its managed workflow engine slots into more than ten popular agent frameworks — LangGraph, Microsoft Agent Framework, Google's Agent Development Kit, AWS Strands, OpenAI Agents SDK, Claude Managed Agents, CrewAI, Pydantic AI Agents and a few others — and adds two things most of them lack on their own: durable execution and verifiable execution.
The pitch is refreshingly low-friction. Teams don't tear out their existing setup. They add a Diagrid code package to whatever framework they've already committed to, and their agent workflows suddenly gain the ability to resume from the exact point of failure instead of restarting from scratch or waiting on a human to intervene. Diagrid points to LangGraph as a case in point: it's strong at orchestration but only offers basic checkpointing, meaning developers have historically had to write their own detection and recovery logic. Run the same LangGraph workflow through Catalyst 2.0, and it finishes after a failure with none of that custom code.
The verification layer is arguably the more interesting piece, and it traces back to Dapr 1.18, which came out in June. Every step in a workflow gets cryptographically signed, with execution lineage propagated and the whole thing attestable after the fact. That's not a nice-to-have for security and compliance teams working in regulated industries — it's the difference between claiming an agent did something and proving it, with a chain of custody attached to each decision.
Diagrid says Catalyst 2.0 runs in the cloud, on-premises, or fully air-gapped, and claims performance up to 10 times better than open-source Dapr, enough headroom to handle millions of concurrent agent workflows. CEO Mark Fussell put the framing bluntly: every agent framework released in the past two years has made building agents easier, but none have made trusting them in production easier. CTO Yaron Schneider, who also chairs a workflows working group at the Agentic AI Foundation, describes this as the industry's shift from making models smart to making agent systems trustworthy. German optics maker ZEISS Group is already using it, with an executive there framing Catalyst as a stable base to build on while the underlying models and frameworks keep changing shape.
Diagrid was founded in 2021 by the creators of Dapr and KEDA, and the company remains a lead maintainer of Dapr within the Cloud Native Computing Foundation. It has raised $24.2 million in venture funding, most of it from a 2022 Series A led by Norwest Venture Partners.
My take — AI-written commentary, not fact-checked reporting
The framework wars have been about who can spin up an agent the fastest, and that race has clearly plateaued into a commodity fight. What actually decides whether any of this gets deployed at scale is boring plumbing — recovery, auditability, chain of custody — and Diagrid is smart to bet that enterprises will pay for reliability long before they pay for cleverness. Cryptographic verification of every agent step also feels less like a luxury feature and more like the thing regulators are going to demand anyway, so building it in now rather than bolting it on later looks like the right call.
Read more about this at: SiliconANGLE