What is a Long-Running Workflow?

A long-running workflow is an automated process that executes over an extended period — minutes, hours, or days — often pausing between steps to wait on external events like a human approval, a scheduled time, or a third-party system's response.

Unlike a simple request-response function, a long-running workflow has to survive interruption: the process executing it might restart, the server might redeploy, or the workflow might sit idle for days waiting on a signature, and the system needs to resume exactly where it left off rather than losing state or restarting from scratch. This durability requirement is why long-running workflows are typically implemented with an explicit state machine or a durable-execution framework that checkpoints progress after every step, rather than as an in-memory function call that dies if the process does.

Designing a long-running workflow well means treating every external wait as a first-class state rather than a blocking call — the workflow engine records "waiting for approval" as a durable state and resumes on the approval event, however long that takes, instead of holding open a thread or connection. Observability matters more here than in short-lived automation, since a workflow stuck for three days needs to be easy to inspect and diagnose, not a black box someone eventually notices has stalled.

In practice with Neotask

A Neotask onboarding workflow can span days — waiting on a customer to connect their email, then a scheduled first-blog-post generation task, then a review approval — and the system durably tracks exactly which stage each tenant is in rather than needing the process to stay alive the whole time. If the server restarts mid-workflow, the tenant resumes at the correct step with no lost progress.

Related terms

• Workflow Automation
• Long-Term Memory (AI)
• LLM Routing
• Audit Logging
• Least Privilege Agent Access

Plans

Related terms

• GraphQL - GraphQL is a query language and runtime for APIs that lets a client specify exactly which fields of data it needs in a single request, rather than…
• Data Lake - A data lake is a centralized repository that stores raw data in its native format - structured, semi-structured, or unstructured - without requiring…
• Top-p Sampling - Top-p sampling, also called nucleus sampling, restricts a language model's next-token choice to the smallest set of tokens whose cumulative…
• Idempotent API - An idempotent API is an API endpoint designed so that calling it multiple times with the same input produces the same server-side result as calling…
• Coding Agent - A coding agent is an AI system that can read, write, and modify source code autonomously across multiple steps — planning changes, editing files…
• API Gateway - An API gateway is a single entry point that sits in front of one or more backend services, handling cross-cutting concerns like authentication, rate…

Explore: Integrations · Skills · Solutions · Use cases · Examples · Comparisons · Templates · Blog · Docs