Diagrid Catalyst 2.0 Adds Call-Level Durability and Cryptographic Attestation

Diagrid announced Catalyst 2.0 on 28 July 2026, adding failure recovery and cryptographic workflow attestation to agents built on ten frameworks: LangGraph, Microsoft Agent Framework, Google ADK, Dapr Agents, LangGraph Deep Agents, AWS Strands, OpenAI Agents SDK, Claude Managed Agents, CrewAI, and Pydantic AI. The release targets a concrete operational failure — a long multi-step agent run that fails late in its sequence forces a full retry, re-paying for every model and tool call already completed. By representing each model and tool call as a durable workflow activity, Catalyst can resume a run from the exact call that failed rather than from the last checkpoint boundary.

For architects weighing where pipeline architecture and orchestration decisions belong in production agent systems, Catalyst's design answer is explicit: durability and attestation live in the infrastructure layer, outside and beneath whatever agent framework the application uses.

Durable Execution

Developers add a Diagrid package to an existing agent application; Catalyst then wraps model and tool invocations as workflow activities backed by Dapr's execution engine. Durable agents are built with a Python SDK, while workflow SDK support extends to .NET, Go, Java, JavaScript, and Python; open-source Dapr 1.18 SDKs additionally cover Rust. Diagrid states Catalyst runs in cloud, on-premise, and air-gapped environments. The Enterprise Server edition, priced on a custom-quoted basis, covers the latter two. Cloud deployments use a tiered commercial model: a free cloud tier, dedicated-cloud, and bring-your-own-cloud plans sized by concurrent workflow count.

Diagrid claims Catalyst can deliver up to ten times the performance of open-source Dapr and support millions of concurrent agent workflows. The announcement specifies neither which metric the ten-times figure describes nor the workload, hardware, or Dapr configuration used as the baseline, so the claim cannot be independently assessed.

Cryptographic Attestation via Dapr 1.18

The verification model derives from Dapr 1.18, which groups the capability into three named features: Workflow History Signing, Workflow History Propagation, and Workflow Attestation. 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. When workflow state is loaded, the full chain is verified, making deleted, reordered, or modified history detectable. Receivers validate each signed chunk against the Dapr Sentry trust anchor, enabling verification outside the originating application — which is what the Propagation and Attestation capabilities carry across workflow and service boundaries.

Several operational constraints matter for rollout. Signing is disabled by default, gated behind the WorkflowHistorySigning feature flag and dependent on mTLS — daprd refuses to start if signing is enabled with mTLS off. Signing is a one-way decision per workflow: existing history cannot be retroactively signed, and enabling signing mid-run on an already-running unsigned workflow produces a verification error. In-flight unsigned workflows must complete or be purged before the flag is enabled more broadly.

As agent control layers increasingly lack shared contracts across providers, cryptographic attestation of workflow history offers a concrete mechanism for establishing post-hoc provenance — though it proves the integrity of recorded history, not the correctness of decisions made or the accuracy of tool outputs.

Competitive Positioning

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.

Named early adoption is limited. ZEISS Group is cited as the sole early user, with Wendelin Niesl, its head of end-to-end core application engineering, stating that Catalyst provides "the stable foundation we can rely on" to build "a sustainable, durable, and resilient platform for both AI and traditional workloads." Diagrid CTO and co-founder Yaron Schneider, who chairs the Agentic AI Foundation's Workflows and Process Integration Working Group, framed the release around organisations needing "proof of what happened" when agents call tools or delegate work.

Practitioners evaluating Catalyst should independently assess retry behaviour for non-idempotent tools, storage and latency overhead added by signing, certificate rotation procedures, and which capabilities belong to open-source Dapr 1.18 versus the commercial Catalyst layer. As autonomy defaults in agent systems raise the stakes for verifiable behaviour, the boundary between history integrity and decision-level auditability remains a gap that attestation alone cannot close.