DBOS
Durable execution for AI agents and apps — checkpoint every step to Postgres, resume on crash.
About
DBOS is an open-source library that makes workflows and queues durable by persisting program state in Postgres. When an app or AI agent crashes mid-run, DBOS retrieves unfinished workflows from the database and resumes exactly where it left off — no separate orchestration server to host, and each step's output commits in the same transaction as the rows it wrote. It supports TypeScript, Python, Java, and Go.
The technology comes from joint MIT-Stanford research led by co-founders Turing Award laureate Mike Stonebraker and Databricks co-founder Matei Zaharia. For agent teams, DBOS shipped a @dbos-inc/vercel-ai package that makes Vercel AI SDK agents durable, plus DBOSify, a drop-in replacement for Temporal Python that runs entirely on Postgres. Durable execution is increasingly the difference between a demo agent and a production one that survives real infrastructure failures.
What it can do
- Durable workflows with automatic crash recovery
- Durable queues, cron scheduling, and notifications
- Exactly-once semantics via Postgres transactions
- Vercel AI SDK integration for durable agents
Setup
Install the DBOS Transact library (npm or pip), point it at a Postgres database, and annotate functions as workflows and steps.
Prerequisites
A Postgres-compatible database; Node.js or Python 3.11+.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 60; Trending score 78/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.