Diagrid announced Catalyst 2.0 on 28 July 2026, adding failure recovery and cryptographic verification to agents built with frameworks including LangGraph, Microsoft Agent Framework, Google ADK and Dapr Agents.
Ten framework integrations ship in the release, also covering LangGraph Deep Agents, AWS Strands, OpenAI Agents SDK, Claude Managed Agents, CrewAI and Pydantic AI.
Developers add a Diagrid package to an existing agent application. Diagrid says Catalyst then represents model and tool calls as durable workflow activities, allowing interrupted runs to resume without repeating completed work.
The failure mode this targets is a familiar one: a long multi-step run fails late in the sequence and, without checkpointing at call granularity, the retry re-pays for every model call already completed. Diagrid says Catalyst runs in cloud, on-premise and air-gapped environments. Durable agents are built with a Python SDK, while Catalyst's documentation lists workflow SDK support for .NET, Go, Java, JavaScript and Python; the open-source Dapr 1.18 SDKs also include Rust.
The verification model comes from Dapr 1.18. Dapr hashes batches of workflow-history events, links each digest to the preceding signature and signs the result using the Dapr sidecar's SPIFFE-based identity. The chain is checked when workflow state is loaded, making deleted, reordered or modified history detectable, and receivers validate each signed chunk against the Dapr Sentry trust anchor, so histories can be verified outside the application that produced them. Dapr groups the work as three capabilities: Workflow History Signing, Workflow History Propagation and Workflow Attestation, the latter two carrying verified execution context across workflow and service boundaries.
Teams should note that signing is disabled by default in Dapr 1.18, gated behind the WorkflowHistorySigning
feature flag and dependent on mTLS, with daprd refusing to start if signing is enabled and mTLS is off. It is also a one-way decision per workflow: there is no retroactive signing of existing history, and toggling signing for a workflow already running is a verification error, so in-flight unsigned workflows must complete or be purged before enabling it more widely.
Yaron Schneider, Diagrid co-founder and CTO and chair of the Agentic AI Foundation's Workflows and Process Integration Working Group, said organisations need "proof of what happened" when agents call tools or delegate work.
Named early adoption is limited so far: optics manufacturer ZEISS Group is cited as an early user, with Wendelin Niesl, its head of end-to-end core application engineering, saying in the announcement that in "a rapidly evolving landscape of AI models and frameworks, Catalyst provides the stable foundation we can rely on", enabling the company to build "a sustainable, durable, and resilient platform for both AI and traditional workloads".
Catalyst is neither the first durability option for these frameworks nor a new entrant to a thin market. LangGraph persistence records checkpoints at graph superstep boundaries, supports resuming from a successful checkpoint and offers persistent task execution through its Agent Server, while Temporal and Restate provide replay or journal-based execution for long-running applications and can already host agent workflows. What sets Catalyst apart is a single Dapr-based recovery and attestation model spanning several frameworks, applying durability to individual model and tool calls rather than graph boundaries. The architectural decision is therefore where teams want recovery, identity, audit evidence and framework integration to reside.
Diagrid claims Catalyst can deliver up to ten times the performance of open-source Dapr and support millions of concurrent agent workflows. The announcement does not say whether that multiple refers to throughput, workflow starts per second or latency, nor which workload, hardware or Dapr configuration it was measured against, so the comparison cannot be independently assessed. Commercially, Diagrid publishes pricing spanning a free cloud tier, dedicated-cloud and bring-your-own-cloud plans sized by concurrent workflows, and a custom-quoted Enterprise Server edition for on-premise and air-gapped deployments.
Cryptographic attestation proves the integrity and provenance of recorded history, not that an agent made a correct decision, a tool returned accurate data, or every external side effect was captured. Practitioners should also assess retry handling for non-idempotent tools, storage and latency overhead, certificate rotation, independent verification, and which capabilities belong to open-source Dapr versus commercial Catalyst.