- Published on
Diagrid Catalyst 2.0 Adds Durable and Verifiable Execution for AI Agents
- Authors
- Name
- DeepHorus
Diagrid has launched Catalyst 2.0, a significant update to its agent development platform that introduces durable and verifiable execution for AI agents. Released on July 28, 2026, the update directly addresses a common and costly problem in long-running AI processes: failure recovery and the cryptographic verification of agent actions. This move aims to provide a more robust infrastructure for complex, multi-step AI workflows.
The release integrates with ten major agent frameworks, including LangGraph, Microsoft Agent Framework, Google ADK, and Dapr Agents. Developers can enhance existing applications by adding a Diagrid package. The system then models model and tool calls as durable workflow activities, allowing an interrupted process to resume from the point of failure without redoing completed steps, a key feature for cost and time efficiency.
Catalyst 2.0 operates across cloud, on-premise, and air-gapped environments. It leverages the open-source Dapr 1.18 runtime to provide cryptographic verification. This verification model uses SPIFFE-based identities to sign and chain workflow history events, making any tamperingâsuch as deletion or reorderingâdetectable both within and outside the original application.
The introduction of durable agents is supported by a Python SDK, while Catalystâs broader workflow capabilities extend to .NET, Go, Java, JavaScript, and Rust through Dapr SDKs. This multi-language support broadens its potential use across diverse development teams and existing technology stacks. The platform's architecture is designed to handle the specific failure modes common in lengthy, sequential AI operations.
Adoption is in its early stages, with optics manufacturer ZEISS Group cited as an initial user. Wendelin Niesl, ZEISSâs head of core application engineering, praised Catalyst for providing a "stable foundation" for building a resilient platform for both AI and traditional workloads. This endorsement highlights the product's potential for enterprise-grade reliability in production environments.
While competitors like LangGraph offer persistence and Temporal provides replay capabilities, Diagrid positions Catalyst as a unified solution. Its key differentiator is a single recovery and attestation model spanning multiple frameworks, applying durability at the individual call level rather than at higher graph boundaries. This approach offers a distinct architectural choice for managing state and audit evidence.
Diagrid claims Catalyst can deliver up to ten times the performance of open-source Dapr and support millions of concurrent workflows. However, the announcement lacks specifics on the metrics, hardware, or configurations used for this comparison. The company emphasizes that cryptographic signing is a one-way, opt-in feature requiring mTLS, which must be considered during deployment planning.
