Temporal

Temporal is an open-source durable execution platform where developers write workflows as code in native SDKs. State is captured at every step, so processes retry and resume after failures. Self-host it or use Temporal Cloud.

Temporal is an open-source durable execution platform for developers who build backend systems that must finish what they start. Instead of wiring together retries, queues, timers and hand-rolled state machines, engineers write business logic as ordinary code in a Temporal Workflow, and the platform records state at each step so execution can resume exactly where it stopped after a crash, timeout or network fault. The programming model has two main pieces. A Workflow holds the business logic and has a durable, fault-tolerant running state, which means it can be recovered, replayed or paused at any point. Activities wrap the failure-prone work, such as calling an API or sending an email, and retry automatically. The Temporal Service persists application state and supplies built-in retries, task queues, signals and timers. Native SDKs are offered for Python, Go, TypeScript, Ruby, .NET, Java and PHP, so teams stay in the language they already use. A visibility view shows the exact state of each Workflow execution, which reduces the need to dig through logs. The homepage emphasizes AI workloads: agents, MCP servers and training pipelines that run long and touch unreliable services. Other listed patterns include human-in-the-loop approvals, Saga-style compensation, long-running processes lasting days or months, order fulfillment, durable ledgers, CI/CD, customer onboarding and DAG orchestration. Named adopters include OpenAI, NVIDIA, Salesforce, Twilio, Descript, Cloudflare and GitLab. Teams can host the Temporal Service themselves from the open-source project or use the managed Temporal Cloud. Compared with general job queues or workflow schedulers, Temporal asks developers to adopt its programming model and deterministic Workflow code, but in return removes much of the reconciliation logic around distributed failures. It suits engineering teams running business-critical, multi-step processes rather than casual automation users looking for no-code connectors.