Summary
On October 15, 2026, Vercel moved Daemons to public beta, letting teams run long-running processes that pull and execute work without an HTTP server. Workers can consume from Vercel Queues, databases, or systems like Celery, BullMQ, RabbitMQ, Kafka, and Temporal, running as containers on Fluid compute in the same project as the app.
What changed
Vercel Daemons entered public beta, adding long-running queue consumers, workflow workers, and self-hosted CI runners that run as containers on Fluid compute. Workers are configured via a daemons entry in vercel.toml/ts/json, built from a Dockerfile, kept running at a chosen instance count per region, and auto-restarted on failure, with output in project Logs and CPU/memory metrics via Observability Plus.
Why it matters
Daemons extends Vercel beyond request/response serverless into always-on background compute, a prerequisite for agentic and workflow workloads that outlive a single HTTP request. It reduces the need to run separate infrastructure for queues and workers, keeping more of the agent stack on Vercel.
Evidence excerpt
Run queue consumers, workflow workers, and self-hosted CI runners on Vercel. Daemons lets you run long-running processes that pull and execute work without requiring an HTTP server.