Temporal
What Temporal is, how workflows, activities and workers fit together, what Temporal Cloud costs as of October 2026, and when self-hosting makes sense.
Open-source (MIT) durable execution platform: write workflows as code and Temporal guarantees they run to completion.
Best for
Engineering teams building long-running, failure-prone business processes in code (payments, provisioning, order flows, AI agents).
Not great for
Ops and GTM teams that want a visual builder with SaaS connectors.
Verdict
The reference implementation of durable execution. Excellent if your team writes code; the wrong tool if you need no-code automation.
Pros
- + Workflows survive crashes, deploys and outages without hand-written state machines
- + Mature SDKs in many languages
- + Open-source server under the MIT license
Cons
- − Code-only: no visual builder or SaaS connector catalog
- − Determinism rules for workflow code take time to learn
- − Self-hosting means running the server plus a database yourself
Key features
- • Durable workflow execution with automatic state persistence
- • Activities with configurable retries and timeouts
- • Signals, queries and updates for interacting with running workflows
- • Durable timers and schedules
- • Workflow versioning for safe code changes
- • Web UI for inspecting workflow history
- • Self-hosted server (MIT) or managed Temporal Cloud
What is Temporal?
Temporal is an open-source durable execution platform. You write a workflow as ordinary code, and Temporal guarantees that it runs to completion, even if the process running it crashes, the machine restarts, or a downstream API is down for an hour. The company behind it, Temporal Technologies, was founded in 2019 by the engineers who built Uber’s Cadence, and the Temporal server is published on GitHub under the MIT license.
When people search “what is Temporal” in a software context, this is the answer: it is infrastructure for reliable, long-running processes, not a no-code automation app.
How Temporal works
A Temporal application has three moving parts:
- Workflows are your business logic written in code (Go, Java, TypeScript, Python, .NET, PHP or Ruby). A workflow execution runs exactly once and to completion. Its state is reconstructed from an event history, which is why workflow code must be deterministic: no direct network calls, random numbers or wall-clock reads inside the workflow itself.
- Activities are the side-effecting steps: calling Stripe, writing to a database, sending an email. Each activity gets its own retry policy and timeouts, so a flaky API is retried automatically instead of failing the whole process.
- Workers are processes you host. They poll the Temporal Service for tasks and execute your workflow and activity code. The Temporal Service itself is a Go server backed by a database (or Temporal Cloud, if you do not want to operate it).
Running workflows can be signaled (send it data), queried (read its state) and updated. Durable timers let a workflow sleep for days, which is how teams implement trial expiry, dunning and approval timeouts without cron jobs.
Temporal pricing
As of October 2026:
- Self-hosted: the server is MIT-licensed and free to run. Your cost is infrastructure (the server cluster plus Postgres, MySQL or Cassandra) and the engineering time to operate and upgrade it.
- Temporal Cloud, Pay-As-You-Go: starts at $50 per million actions with no base fee, plus storage billed per GB-hour for active and retained workflow history. Actions are billable operations such as starting a workflow, running an activity, a timer or a signal.
- Support add-ons: Business starts at $500/month (includes 2.5M actions); Enterprise and Mission Critical are quoted by sales.
Check temporal.io/pricing for current numbers, because action-based pricing changes with your workflow design: a workflow with 20 activities costs more than one with 3.
Who Temporal is for (and not for)
Good fit: backend teams with processes that span minutes to months and must not lose state: payment and order orchestration, infrastructure provisioning, onboarding sequences, AI agent loops with tool calls and human approval.
Poor fit: ops, RevOps or marketing teams who want to connect HubSpot to Slack. Temporal has no visual builder and no connector catalog; every integration is code you write inside an activity.
Temporal alternatives
- Restate: a newer durable execution runtime shipped as a single binary.
- Inngest: event-driven durable functions, compared head-to-head in Inngest vs Temporal.
- Prefect and Airflow if your workflows are mostly data pipelines.
- See the workflow engine entry for the wider category.
How Temporal compares to Automatite
They solve different problems. Temporal gives engineers a programming model for durable code; Automatite is a workflow automation platform with a visual workflow builder, triggers, SaaS actions, and code steps for the parts that need custom logic. If your team writes Go or Java services and needs exactly-once orchestration, use Temporal. If your ops team needs to automate processes across business apps without owning workers, Automatite is the closer fit.
FAQ
Is Temporal free?
The Temporal server is open source under the MIT license, so self-hosting is free apart from infrastructure and operations. Temporal Cloud is paid and metered by actions and storage.
What does Temporal do?
It runs workflow code durably: it persists every step’s result, retries failed activities, and resumes execution after crashes so long-running processes complete reliably.
Is Temporal a no-code tool?
No. Temporal is a developer platform; workflows are written with its SDKs.