Diagrid Catalyst

Diagrid Catalyst is a coding tool. Diagrid Catalyst adds durable workflows, cryptographic identity, mTLS and tamper-evident tracing beneath AI agents built on frameworks such as LangGraph, CrewAI and Google ADK. It targets enterprise teams running agents and MCP servers in production.

Diagrid Catalyst is a managed runtime layer that sits underneath AI agents and makes them survive failures. Instead of replacing the agent framework a team already uses, it wraps the agent in durable workflows so that a crash, deploy or restart resumes the run where it stopped rather than starting over. The homepage lists support for LangGraph, CrewAI, Google ADK, Strands, OpenAI Agents, Microsoft Agent Framework, Spring AI, Dapr Agents, Pydantic AI and others, plus plain applications that use no agent framework at all. In practice, a developer defines an agent as usual and then hands it to a Dapr-based workflow runner, which the site presents as a change of only a few lines. From there Catalyst provides the pieces production agents tend to lack: persisted state, retries and checkpoints, session management, pub/sub messaging for agent-to-agent coordination, and service discovery. Every agent and MCP server receives a cryptographic identity, communication uses mutual TLS, and each step can be signed into a tamper-evident record that supports tracing and deterministic replay for audit work. The platform is built on open-source projects including Dapr, SPIFFE, CloudEvents, KEDA and OpenTelemetry. A single API fronts message brokers, databases, secret stores, configuration stores and LLM providers, so teams can swap Kafka for RabbitMQ or one model provider for another without rewriting application code. Multi-region failover routes traffic between regions and clouds during outages. Deployment is either Diagrid-hosted (cloud or dedicated cluster) or bring-your-own-cloud, and a self-hosted enterprise option exists. The audience is enterprise AI teams, platform engineers and security or compliance groups that need agents to run reliably and be provable after the fact. Compared with building these capabilities by hand, or adopting a workflow-only engine that lacks agent identity and attestation, Catalyst positions itself as a single integrated system. Teams building quick prototypes will find it more infrastructure than they need.