DuraGraph

Mission

DuraGraph provides an open, extensible orchestration layer for AI and workflow automation. Our mission is to enable reliable, observable, and maintainable pipelines that feel natural for developers—bringing the power of LangGraph Cloud to self-hosted and enterprise environments.

Compatibility with LangGraph Cloud

DuraGraph implements an API shim compatible with LangGraph Cloud. This means:

• Tools, clients, and SDKs built for LangGraph Cloud can run against DuraGraph.
• OpenAPI contracts are aligned for seamless integration.
• Migration and hybrid deployments are supported.

Why Temporal?

We selected Temporal for the control plane and core orchestration engine because:

• It provides stateful, durable workflow management.
• Built-in retries, error handling, and compensation logic.
• Proven scalability for high throughput workloads.
• Rich observability and workflow introspection.

This ensures fault tolerance and deterministic execution for AI-driven processes.

Target Users

• AI platform teams building internal orchestration layers.
• Enterprises that require compliance, security, and observability in AI workflows.
• Open-source contributors looking for a community-driven orchestration stack.
• Researchers who want to run reproducible LangGraph experiments locally.

Non-Goals

• Providing a custom LLM runtime (we integrate with existing providers).
• Replacing Temporal itself (we build on top of it).
• Becoming a full-featured cloud platform (our scope is orchestration, not hosting).

Success Metrics

We measure success by:

• Adoption: # of teams using DuraGraph in production.
• Compatibility: seamless drop-in replacement for LangGraph Cloud APIs.
• Reliability: demonstrated ability to recover from failures without data loss.
• Community growth: contributions and RFCs from external developers.

What's Included

flowchart LR
  client["Client SDKs / LangGraph Cloud clients"]
  shim[API Shim]
  translator[Translator / IR]
  temporal[Temporal Control Plane]
  workers[Worker Adapters]
  storage[(Storage)]
  observability[(Observability / Metrics / Tracing)]

  client --> shim --> translator --> temporal --> workers
  workers --> storage
  workers --> observability