# Ship Reliable AI Agents to Production Stop building infrastructure. Start shipping agents that never fail. - [Build Your First Production-Ready Agent](https://docs.diagrid.io/getting-started/quickstarts/ai-agents) ## What is Diagrid Catalyst? Diagrid Catalyst is an **agentic durable execution platform** for governing, securing, and running agents and workflows in production. It gives every workload — an agent, a workflow, an MCP server, or an application — a verifiable identity, durable execution, and policy-driven access, without writing security or infrastructure code in your application. Survive every failure, authorize every interaction. Catalyst works with your existing agent frameworks and MCP servers to add self-recovering execution, zero-trust security, and cryptographic identity to every agent and tool. It automatically resumes agents and workflows from crashes, deploys, and restarts with no lost progress, protects critical resources with policy-based access control, lowers LLM and infrastructure costs through reliable execution, and accelerates the path from prototype to production. ![Diagrid Catalyst sits between your agent and app SDKs and your infrastructure, providing durable agents and workflows, authentication and authorization, and governance](https://docs.diagrid.io/img/catalyst/catalyst-overview-light.png) **[Durable execution](https://docs.diagrid.io/concepts/durable-execution)** keeps work running through failure; **[verifiable execution](https://docs.diagrid.io/concepts/verifiable-execution)** lets you prove how it ran. Catalyst cryptographically signs each step of a workflow's history to produce a tamper-evident record that anyone can independently verify — a chain of custody for autonomous systems. When an agent delegates work or calls a tool, you can prove not just *who* made a request, but *how* the execution reached it: a bank can confirm a payment came from an approved workflow, and a regulated team can show a process ran every required step. Catalyst is built on open-source [Dapr](https://docs.diagrid.io/dapr-open-source), a graduated CNCF project, and open standards, so enterprises get full sovereignty over their AI infrastructure across cloud, on-premises, and hybrid environments — built for production, not demos. You deploy Catalyst into your own cloud or on-premises and run it as a shared, governed platform for your teams. Review the [Catalyst hosting model](https://docs.diagrid.io/operate/hosting) to see how the control plane and data plane map onto your environment, and the [connect your apps guide](https://docs.diagrid.io/develop/connect) for the minimal changes needed to bring existing apps onto Catalyst. With Catalyst, your code stays focused on business logic instead of re-inventing distributed systems. It works with any language, any framework, and any cloud — including production-ready integrations with [Dapr Agents](https://docs.diagrid.io/develop/agents/dapr-agents), [CrewAI](https://docs.diagrid.io/develop/agents/crewai), [LangGraph](https://docs.diagrid.io/develop/agents/langgraph), [Strands](https://docs.diagrid.io/develop/agents/strands), [OpenAI Agents](https://docs.diagrid.io/develop/agents/openai), [Google ADK](https://docs.diagrid.io/develop/agents/adk), [Pydantic AI](https://docs.diagrid.io/develop/agents/pydantic-ai),[Microsoft Agent Framework](https://docs.diagrid.io/develop/agents/microsoft), [Mastra](https://docs.diagrid.io/develop/agents/mastra) and [Claude Agents](https://docs.diagrid.io/develop/agents/claude) ## Start Building Pick your path. Build a production-ready agent or workflow in 5 minutes. Each card below links to a hands-on [quickstart](https://docs.diagrid.io/getting-started/quickstarts) — follow along in your language of choice, or jump to the deeper [agent framework guides](https://docs.diagrid.io/develop/agents), [workflow SDK guides](https://docs.diagrid.io/develop/workflows), and [Dapr API SDK guides](https://docs.diagrid.io/develop/dapr-apis) once you have the basics working. - [AI Agents](https://docs.diagrid.io/getting-started/quickstarts/ai-agents) — Build agents that survive crashes, coordinate reliably, and remember everything. Works with Dapr Agents, CrewAI, LangGraph, Strands, OpenAI Agents, Google ADK, Pydantic AI, and more. - [Workflows](https://docs.diagrid.io/getting-started/quickstarts/workflow) — Orchestrate complex processes that run for hours or days. Automatic retries, checkpointing, and recovery built in. - [Dapr APIs](https://docs.diagrid.io/getting-started/quickstarts/dapr-apis) — Build distributed applications using Dapr's building blocks — state management, pub/sub messaging, service invocation, jobs, and more. --- Ready to move beyond the basics? Scaffold a complete Dapr Workflow project from a diagram with [Workflow Composer](https://docs.diagrid.io/develop/workflows/workflow-composer), connect agents to tools through [MCP on Catalyst](https://docs.diagrid.io/develop/mcp), and inspect running workflows and agents locally using the [Dapr Dev Dashboard](https://docs.diagrid.io/develop/local-development/dev-dashboard). ## Learn the Fundamentals - [Dapr Building Blocks](https://docs.diagrid.io/dapr-open-source) — State, pub/sub, workflows, service invocation—pre-built APIs that replace thousands of lines of infrastructure code. - [Catalyst Platform](https://docs.diagrid.io/operate) — Managed sidecars, centralized control plane, unified observability. You build apps, we run the infrastructure. - [Dapr & AI University](https://www.diagrid.io/dapr-university) — Free, self-paced courses. Learn Dapr fundamentals, workflows, and agent patterns through hands-on coding. New to Dapr and Catalyst? These two free, hands-on tracks are the fastest way in. **Dapr & AI University — Start here:** [Dapr 101: State Management, Service Invocation, and Pub/Sub APIs](https://www.diagrid.io/university/dapr-101?utm_source=docs&utm_medium=cta&utm_campaign=dapr-101) Learn Dapr's core building blocks — state management, service invocation, and pub/sub — through hands-on coding. Beginner · 45 min · .NET, Python, Java, JavaScript **Dapr & AI University — Then run it on Catalyst:** [Running Dapr applications with Diagrid Catalyst](https://www.diagrid.io/university/catalyst-101?utm_source=docs&utm_medium=cta&utm_campaign=catalyst-101) Run Dapr apps on Catalyst's managed pub/sub, KV store, and workflow engine — no infrastructure to operate. Beginner · 30 min · .NET, Python, Java, JavaScript --- ## Deploy Your Way Catalyst has four deployment models. They differ in where the Catalyst data plane runs and who operates it — [compare them side by side](https://docs.diagrid.io/operate/hosting). - [Catalyst Cloud](https://catalyst.diagrid.io/) — A fully managed, serverless platform for durable workflows, agents, and MCP servers. Free to start. - [Catalyst Dedicated Cloud](https://docs.diagrid.io/operate/hosting/dedicated-cloud) — A single-tenant deployment sized for production, provisioned and operated for you on AWS or Azure. - [Catalyst BYOC](https://docs.diagrid.io/operate/hosting/byoc) — The Catalyst data plane runs in your own cloud, so application data never leaves your network. - [Catalyst Enterprise Server](https://docs.diagrid.io/operate/hosting/enterprise-server) — Control plane and data plane both run on your infrastructure, with no external connectivity required.