Side by side
| DurbarAI | Langfuse | |
|---|---|---|
| What it is | A runtime that runs agents and governs them. | Tracing, evals, datasets, prompt management and cost tracking. |
| License | Not public yet. Open-source release planned; license not decided. | Core is MIT. Some self-hosted features need an enterprise license key. |
| Human approval of tool calls | Built in, with a named approver and an audited decision. | Not found. Annotation queues let people score traces, not approve actions. |
| Release gate | Every promotion is scored against the live version and blocked if it falls short. | Evals and experiments. A gate on eval scores was not found; protected labels limit who can change a prompt label. |
| Audit | Per-event HMAC hash chain, Ed25519-signed checkpoints, one-click verify. | Audit logs, enterprise license when self-hosted. |
| Self-hosting | Docker Compose or Kubernetes. Images built from source. | Yes; core features free under MIT. |
When Langfuse is the better fit
- You need tracing and evals across many apps and frameworks, not a runtime.
- You want a mature open-source project with a public repository today.
When DurbarAI is the better fit
- You need agents that wait for a person before they change something.
- You want releases blocked, not only measured, when a new version does worse.
Using both
DurbarAI writes OpenTelemetry GenAI spans, and Langfuse accepts OpenTelemetry traces, so Langfuse can be where you look at DurbarAI runs.
Sources: facts about Langfuse come from Langfuse pricing, Langfuse self-hosting, Langfuse license key features, 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.