Braintrust is a commercial AI evaluation and observability platform; self-hosting is Enterprise-only and keeps the control plane with Braintrust. Pages here cover replacing it with self-hosted Langfuse or Opik.
Comparing Braintrust with self-hosted Langfuse for tracing, evaluations, datasets and prompt management: Braintrust's processed-data and score pricing against Langfuse's open-core license, which Langfuse features need an Enterprise key, and who operates ClickHouse, PostgreSQL and storage.
Comparing Braintrust with Opik, whose Apache-2.0 repository includes the full platform with tracing, evaluations and prompt management: evaluation workflow, experiment comparison, self-hosting effort on ClickHouse and MySQL, and when Braintrust's managed control plane is worth paying for.
Moving from Braintrust to self-hosted Langfuse: exporting logs, datasets and experiments as JSON or Parquet through BTQL in pages that respect the 30-second timeout, mapping scores and metadata, re-instrumenting applications, and rebuilding evaluators and CI gates.
Moving Braintrust experiments, datasets and traces to Opik: export formats, recreating datasets and LLM-as-a-judge metrics, switching SDK instrumentation or OpenTelemetry exporters, and running both systems in parallel until evaluation results match.
Cost of ownership by component: Braintrust Pro at $249 a month plus processed-data, score and retention overages, or an Enterprise contract for self-hosting, against Langfuse's open-source or Enterprise license, ClickHouse, PostgreSQL and object storage, and the people who run them.
What Braintrust lets you export before a contract ends: logs, datasets and experiments from the UI as JSON or CSV, JSON or Parquet from the BTQL API, the permissions the export button needs, the 30-second request timeout and paging, and what to verify before deleting the project.
Braintrust's Enterprise self-hosting keeps the web UI, authentication and metadata in Braintrust's control plane while your cloud holds traces and datasets. This page compares that split with Langfuse or Opik run entirely in-house, for security reviews, residency and network isolation.
Replacing Braintrust's experiments, scorers and dataset-driven evaluations: promptfoo for evaluations run locally and in CI, Opik and Langfuse experiments and LLM-as-a-judge metrics for tracked results, and how to carry scoring functions and baselines across.
Teams using Braintrust's hosted proxy for one API across providers and response caching can deploy the MIT-licensed proxy themselves or move to LiteLLM, which adds virtual keys, budgets, fallbacks and logging. This page compares the two and plans the switch of base URLs and keys.
Tell us what you're working on, whether it's a deployment, an audit, a security test or a cyber range. You'll speak with an engineer who can help you scope it.