Side by side
| DurbarAI | LangSmith | |
|---|---|---|
| What it is | A runtime that runs agents and governs them. | Tracing, evals, prompt tooling and agent deployment. |
| Self-hosting | The whole product, on Docker Compose or Kubernetes. | An add-on to the Enterprise plan. |
| Human approval | Built in: gated tools pause the run, the approver decides in an inbox, the decision is audited. | LangGraph interrupt() pauses a run and resumes it with a decision. The inbox, identity and record are yours to build. |
| Release gate | Every promotion is scored against the live version and blocked if it falls short. | Online and offline evals. A gate that blocks a deployment on eval results was not found on its public pages. |
| Audit | Per-event HMAC hash chain, Ed25519-signed checkpoints, one-click verify. | Audit logs on Enterprise plans, described as tamper-resistant, with 400-day retention. |
| Stop button | Halts new and in-flight runs at the next step. | Not found on its public pages. |
| Published prices | None yet. | Developer $0 a seat, Plus $39 a seat, Enterprise custom. |
When LangSmith is the better fit
- You want hosted tracing and evals with the least setup.
- Your team already lives in the LangChain ecosystem and wants LangChain's own tooling.
When DurbarAI is the better fit
- Agents must run on your servers, without an enterprise contract to get there.
- A risk or security reviewer has to sign off before agents take real actions.
- You want promotion blocked when a new version does worse than the live one.
Using both
DurbarAI writes OpenTelemetry GenAI spans, so its runs can go to whichever tracing backend you already use.
Sources: facts about LangSmith come from LangChain pricing, Self-hosted LangSmith, LangSmith audit logs, LangGraph interrupts, checked on 7 November 2026. "Not found" means we did not find it there, not that it does not exist. DurbarAI's own gaps are on what DurbarAI cannot do yet. Spot an error? Tell us and we will fix it.