Trigger.dev

Trigger.dev is an open source background jobs framework that enables developers to build and deploy fully-managed AI agents and workflows in TypeScript. The platform handles long-running tasks with automatic retries, queuing, observability, and elastic scaling, eliminating timeouts and infrastructure management concerns. It provides a complete solution for offloading async AI tasks to managed infrastructure while supporting human-in-the-loop functionality and real-time streaming to frontends.

• No Timeouts - Write simple, reliable code that runs as long as needed without hitting execution limits, perfect for complex AI agent workflows and long-running processes.
• Automatic Retries - Configure retry policies with exponential backoff, randomization, and custom attempt limits to ensure task resilience and reliability.
• Concurrency & Queues - Control how many tasks run simultaneously with built-in queue management and concurrency limits that only apply when code is actively executing.
• Realtime API - Display task status, progress, and metadata in your frontend using React hooks, with support for streaming LLM responses directly to users.
• Build Extensions - Customize your build process with extensions for Python, Prisma, Puppeteer, FFmpeg, and more, enabling full runtime freedom.
• Observability & Monitoring - Watch tasks execute in real-time with trace views, logging, advanced filtering, and alerts via email, Slack, or webhooks.
• Scheduled Tasks - Create durable cron schedules that run reliably without timeout concerns for recurring background jobs.
• Human-in-the-Loop - Pause runs until tokens are completed via approval workflows using the wait-for-token functionality.

To get started, install the Trigger.dev CLI and SDK in your existing codebase, write tasks in a /trigger folder using plain async TypeScript code, and deploy to Trigger.dev Cloud or self-host on your own infrastructure. The platform supports any JavaScript framework and integrates with existing Node.js SDKs. Tasks are bundled and deployed to managed workers where you only pay for actual execution time.