Last updated: 2026-11-08
DBOS Transact is an MIT licensed durable execution library with SDKs for five language families. Its Python release 3.4.0 shipped on 28 October 2026. Workflow state lives in Postgres, so no orchestration server is needed, and the commercial Conductor console adds recovery, alerts and an MCP server.
About DBOS
DBOS is a durable execution library for backend services and AI agents. You add annotations to ordinary functions, and the library writes each step's result to Postgres so a crashed or restarted process resumes from the last completed step instead of starting over. The company, DBOS, Inc., was co-founded by Postgres creator Mike Stonebraker with Qian Li (CEO) and Peter Kraft (CTO), and reports $8.5M+ raised.
The mechanism is what separates it from Temporal and Inngest. There is no orchestration server to run: the library lives inside your application process and the only infrastructure is a Postgres database (SQLite also works for local use). Workflow state and your application data can sit in the same database, so a step can commit business data and its own checkpoint in one transaction. The trade-off is that throughput is bounded by the Postgres you provide, and the vendor's own benchmarks describe how far a single server scales rather than promising unlimited headroom.
Transact, the library, is MIT licensed and ships SDKs for five language families. Conductor is the commercial layer on top: a workflow management console, automatic recovery of workflows after a crash, retention policies, custom alerts and an MCP server for debugging from a coding agent. It is aimed at engineers who already run Postgres and want durable steps for agent loops, queues, scheduled jobs and payment-style flows.
The agent angle is documented: DBOS has a quickstart for making agent loops durable, and LlamaIndex lists a llama-agents-dbos plugin, so it sits alongside LlamaIndex and LangGraph rather than replacing them. Hosted runners such as Trigger.dev take the opposite approach by running your jobs for you.
Pricing is simple at the top: the library is free, Conductor has Pro and Teams plans, and Enterprise terms are custom for self-hosted Conductor. Every plan beyond the library meters on checkpoints, so a workflow with many small steps costs more than one with a few large ones.
Pricing
The library is free under MIT. Conductor plans are Pro at $99/month, Teams at $499/month and a quoted Enterprise tier, each metered on checkpoints. The vendor describes a free 30-day Conductor trial.
Read from dbos.dev/dbos-pricing on 2026-11-08. That page still lists DBOS Cloud hosting by quote while the vendor llms.txt calls it retired, so it is not counted as a plan.
| Tier | Monthly price | What it includes |
|---|---|---|
| Transact (open source) | Free | MIT licence, Five language families, Postgres you host, No orchestrator |
| Conductor Pro | $99/mo | Seats: 2; apps: 3, Checkpoints: 1M included, Console, recovery, MCP server, Discord community help |
| Conductor Teams | $499/mo | Seats: 10; apps: 10, Checkpoints: 10M included, Alerts and role-based access, SOC 2 Type 2, HIPAA, Slack channel, 2-day SLA |
| Enterprise | Custom | Quoted pricing, Self-hosted Conductor, SSO/SAML, air-gapped, Sharded Postgres, Same-day SLAs |
Key Features
- Postgres-backed checkpoints: Each workflow step result is stored in Postgres, so a crashed process resumes from the last completed step with no separate orchestration server.
- Five language SDKs: Transact ships for Python, TypeScript, Go, Java and Kotlin, and Rust, all under the MIT licence.
- Durable queues: Workflows can be enqueued with concurrency and rate limits, and the vendor reports queue streams reaching 60,000 writes per second on one Postgres server in its own LISTEN/NOTIFY test.
- Conductor workflow console: Conductor lists workflow state and history and lets you cancel, resume or fork a run, with automatic recovery of failed executions.
- Alerts and retention: Conductor Teams adds custom alerts that integrate with Slack and PagerDuty, plus retention policies that offload old workflow data.
- MCP server: Pro and Teams include an MCP server so a coding agent can inspect and debug workflows directly.
Pros
- The only infrastructure is the Postgres you already run, which removes a separate orchestrator cluster that Temporal-style setups need.
- Workflow checkpoints and application data can commit in the same database transaction, which simplifies exactly-once logic for payments and order flows.
- The library is free under MIT with SDKs in five language families, so a team can try it in production code before paying for Conductor.
Cons
- Throughput and storage are bounded by your Postgres, so very high workflow volumes mean sizing and tuning that database yourself.
- Paid plans meter on checkpoints, so step-heavy workflows raise the bill.
- Community support through Discord is the only support on Pro; dedicated Slack and response SLAs start at the Teams plan.
Data Handling
- Compliance
- SOC 2 Type 2 · HIPAA · GDPR
Frequently Asked Questions
What does DBOS actually cost?
The DBOS Transact library costs nothing under the MIT licence. Conductor Pro is $99/month and covers a small team, with overage at $50 per extra million checkpoints. Teams is $499/month with a larger allowance, overage at $40 per million and $50 per added app. Enterprise is quoted.
Can you use DBOS without paying?
Yes. The library has no seat or step limit and runs on any Postgres you provide. Conductor, the console, has a free 30-day trial according to the vendor. After that, Pro is the entry paid plan.
What are DBOS's closest competitors?
Temporal is the established dedicated orchestrator, a fit when you want a separate cluster. Inngest suits teams that want event-driven functions hosted for them. Trigger.dev runs your background jobs on its own infrastructure, which DBOS does not.
How does DBOS compare to Temporal in 2026?
DBOS runs inside your app and stores state in Postgres, while Temporal runs a separate server that your workers call. That makes DBOS lighter to adopt but ties scale to your database. Temporal wins when you need its dedicated service and wider language tooling.
How do you set up DBOS?
Install the SDK for your language, point it at a Postgres connection string, and annotate a workflow and its steps. The vendor quickstart covers a first durable workflow, and an AI quickstart covers agent loops. Conductor is optional and connects through the console sign-in.
Top Alternatives
- Temporal: Pick DBOS if Postgres is enough infrastructure; choose Temporal when you want a dedicated orchestration cluster and its wider tooling.
- Inngest: Choose Inngest when you want event-driven functions hosted for you; pick DBOS when workflow state must stay in your own database.
- Trigger.dev: Pick Trigger.dev to have background jobs run on its infrastructure; choose DBOS to embed durability in services you already deploy.