trigger.dev
Trigger.dev – build and deploy fully‑managed AI agents and workflows
Open-source TypeScript platform for writing long-running background jobs and AI workflows that run without timeouts, with retries, queues, monitoring, and scheduling included.
Trigger.dev is an open-source platform for building and running background jobs and AI workflows written in TypeScript. The README pitches it as a way to write long-running tasks in your own codebase, then deploy them to Trigger.dev's cloud (or self-host) and get retries, queues, monitoring, and automatic scaling included.
The main framing is that other places to run code, like AWS Lambda or Vercel serverless functions, cut off tasks after a short timeout. Trigger.dev says it runs your tasks with no timeout, so you can do things like call multiple LLMs in sequence, process video with FFmpeg, run a browser, or execute Python scripts as part of a job. Tasks are checkpointed, which the README says means they survive restarts and can resume where they left off.
Developers install the SDK and write a task as a normal TypeScript function with a unique id and a run function that receives a typed payload. The example in the README is a hello-world task that just logs a message. Tasks live in your repo, so they go through version control, local testing, and code review like the rest of your code.
The feature list is long. It includes scheduled cron jobs, realtime updates and LLM response streaming to a frontend through React hooks, batch triggering of many runs, structured input and output schemas, waits and waitpoints for pausing a task until a human approves it, concurrency and queue controls, and separate Development, Staging, Preview, and Production environments. Each run has full tracing, logs, tags, metadata, and configurable error alerts.
The README also points to a hosted cloud at cloud.trigger.dev as the quickest way to get started, with self-hosting documented as an alternative. The project is licensed Apache 2.0.
Where it fits
- Run chained LLM calls and AI workflows that exceed the timeout limits of Lambda or Vercel
- Schedule cron jobs and batch-trigger many runs from a TypeScript codebase
- Pause a task until a human approves it using waitpoints, then resume execution
- Stream LLM responses and realtime updates to a frontend via React hooks