Langfuse vs Braintrust: open-source self-hosting or eval-first SaaS
Langfuse is an MIT-licensed open-source platform, owned by ClickHouse since January 2026, that teams can self-host free, while Braintrust is a proprietary eval-first platform with self-hosting only on Enterprise, so Langfuse fits teams that need traces on their own infrastructure and Braintrust fits teams that want a managed eval workflow.
Checked
Langfuse and Braintrust often end up on the same shortlist. Both trace model and tool calls, track cost, store datasets, run evals offline and on live traffic, and accept OpenTelemetry traces. Both appear in Anthropic's January 2026 guide to agent evals.
The split is licence and hosting. Langfuse's core is MIT and the open-source edition self-hosts free. Braintrust is a hosted product with an Apache-2.0 SDK and its own trace database, and self-hosting is part of its Enterprise plan.
Licence and ownership
Langfuse
Acquired by ClickHouse on 16 January 2026, with no planned changes to licensing. Core MIT; ee/ directories under a commercial licence.
Braintrust
Independent, with an $80M Series B led by ICONIQ on 17 February 2026. Platform proprietary; SDK Apache-2.0.
Pricing and hosting
Langfuse Cloud runs from Hobby at $0 (50k units a month, 2 users) to Core at $29, Pro at $199 and Enterprise at $2,499 a month. The open-source edition self-hosts free without limits on core features.
Braintrust runs from Starter at $0 (1 GB processed, 10k scores, 14-day retention) to Pro at $249 a month (5 GB, 50k scores, 30-day retention), with Enterprise custom and the only route to on-prem deployment.
Evals
Langfuse
LLM-as-a-judge and code evaluators, online on live traces and offline on datasets, experiments comparing prompt, model or code changes, a GitHub Action that fails a pull request on regression, annotation queues and score analytics.
Braintrust
Experiments and scorers joined to production traces, the Loop agent, and Topics, generally available since 1 June 2026, which clusters production traces into failure modes. A workflow evals post followed in October 2026.
Tracing, OpenTelemetry and cost
Langfuse accepts OTLP over HTTP (JSON or protobuf) at /api/public/otel and builds its Python and JS/TS SDKs on the OpenTelemetry client. Braintrust accepts OTLP traces at /otel/v1/traces. Langfuse computes cost per generation from model price definitions or ingested values; Braintrust shows estimated LLM cost and tokens on each span, rolled up to parents.
How to decide
If the hard requirement is that traces stay inside the team's own cloud, without an enterprise contract, Langfuse's open-source edition meets it. If the team wants evals to drive releases and is happy on a hosted platform, Braintrust is organised around that loop. Both accept OpenTelemetry, so an agent instrumented once can be trialled on each.
Motiontic's engineers are vendor-neutral on tracing and eval tools. They pick a backend against the client's hosting, data-residency and budget constraints, keep instrumentation portable where the stack allows, build the eval suite from the client's own production failures, and leave the traces, datasets and evals with the client's team.
Langfuse and Braintrust compared for production agents. Prices in USD, as of 1 November 2026
| Langfuse | Braintrust | |
|---|---|---|
| Owner | ClickHouse, since 16 January 2026 | Braintrust (independent) |
| Licence | Core MIT; ee/ under a commercial licence | Proprietary platform; SDK Apache-2.0 |
| Self-hosting | Open-source edition free; Enterprise editions custom | Enterprise only |
| Free tier | Hobby: 50k units, 30-day access, 2 users | Starter: 1 GB, 10k scores, 14-day retention |
| Paid entry | Core $29 a month; Pro $199 a month | Pro $249 a month |
| OpenTelemetry | Accepts OTLP over HTTP; SDKs on the OTel client | Accepts OTLP traces |
| Evals | LLM-as-a-judge, code evaluators, experiments, CI action, annotation queues | Experiments, scorers, CI, Loop agent, Topics clustering |
| Cost tracking | Per generation, from price definitions or ingested costs | Estimated LLM cost per span, rolled up to parents |
Common questions
Is Langfuse or Braintrust open source?
Langfuse's core is MIT open source, with enterprise features under a commercial licence. Braintrust's platform is proprietary; only its SDK is open source, under Apache-2.0.
Can Braintrust be self-hosted like Langfuse?
Only on Braintrust's Enterprise plan, which offers on-prem or hosted deployment. Langfuse's open-source edition can be self-hosted free by any team.
Which is cheaper, Langfuse or Braintrust?
Both have free tiers. Langfuse Cloud paid plans start at $29 a month and Braintrust Pro at $249 a month, but the units differ, so the comparison has to use the team's own trace volume. Self-hosted Langfuse has no licence fee but carries infrastructure and operating cost.
Related
AI agent development
Agents that do real work for users, with the evals, traces and guardrails to keep them in production.
Braintrust vs LangSmith
Two hosted platforms for agent traces and evals, one built around experiments, one around the LangChain stack.
Langfuse alternatives
Seven tracing and eval platforms compared on licence, hosting, strengths and the teams each one fits.
LangSmith vs Langfuse
Hosted platform from LangChain against the open-source, self-hostable platform now owned by ClickHouse.
Eval
Repeatable, scored tests that tell a team whether an AI change made things better or worse.
Sources
- Langfuse: Joining ClickHouse (2026-01-16)
- Langfuse: Cloud pricing
- Langfuse: Self-host pricing
- Langfuse docs: OpenTelemetry
- Langfuse docs: Evaluation overview
- Langfuse docs: Token and cost tracking
- Braintrust: Pricing
- Braintrust: Blog (Series B, Topics GA, workflow evals)
- Braintrust docs: OpenTelemetry
- Braintrust docs: Examine traces
Running agents that have to hold up?
Describe the agent, the systems it touches and what it must never get wrong. The first call is with a lab engineer, who will say plainly whether it is ready for production, and whether Motiontic is the right team.