Both cover tracing, experiments, online scoring, human review and prompt versioning. Opik is the open-source option you can run yourself for free; Braintrust is a hosted product built around experiments, with its Loop agent analysing results for you. The hosting question usually settles it before any feature does.[1][4]
The difference between Braintrust and Opik is mainly licence and hosting: Braintrust is a hosted, proprietary evaluation platform organised around experiments, with self-hosting of its data plane on Enterprise only, while Opik, from Comet, is Apache-2.0 and free to self-host with online evals included. Both trace agents, score production traffic and version prompts. Opik ships guardrails your code can call; Braintrust documents no run-time blocking.
| Decision factor | Braintrust | Opik |
|---|---|---|
| Ownership | Independent; $80M Series B in February 2026 | Comet ML, independent; reports $70M in total funding |
| Licence | Proprietary platform; SDKs and Autoevals are open source | Apache-2.0 for the core platform and SDKs |
| Self-hosting | Enterprise only, and only the data plane; Braintrust runs the control plane | Free; all features except user administration |
| Run-time checks | None documented; scoring runs after logging | Partial: guardrails your code calls; no approval holds |
| Gateway | Hosted gateway in public preview, no SLA | LLM Gateway in beta, for development and testing only |
| Free tier | Starter: 1 GB processed data, 10K scores, 14-day retention | Free Cloud: 25k spans a month, 60-day retention, up to 10 members |
| First paid tier | Pro: $249 a month, unlimited users | Pro Cloud: $19 a month, 100k spans, up to 50 members |
| Capability | Braintrust | Opik | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | ✓ | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | ✓ | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | ✓ | ✓ |
| Online evals on production traffic | ✓ | ✓ | ✓ |
| LLM-as-a-judge scoring | ✓ | ✓ | ✓ |
| Code-based evaluators | ✓ | ✓ | ✓ |
| Human review and annotation | ✓ | ✓ | ✓ |
| Prompt versioning and playground | ✓ | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | — | Partial | ✓ |
| Gateway or proxy in the request path | ✓ | Partial | — |
| Licence and deployment | |||
| Open source | Partial | ✓ | Partial |
| Self-hosting | Partial | ✓ | Partial |
Opik's core platform, covering tracing, evaluation, the UI, backend and SDKs, is Apache-2.0 in a public repository. Self-hosting is free and, according to Comet's documentation, includes every feature except user administration. The local Docker install is described as not production-ready; production deployments run on Kubernetes.
Braintrust is proprietary and hosted. Its SDKs, the Autoevals scorer library and its Terraform modules are open source, but the platform is not. On the Enterprise plan you can run the data plane in your own AWS, GCP or Azure account, or use BYOC, while Braintrust always operates the control plane: the UI, authentication and metadata. If a compliance review requires every component inside your network, Opik is the one of the two that allows it. If it only requires that trace data stays in your cloud account, Braintrust's Enterprise data plane may be enough, and you avoid running the rest yourself.
Braintrust starts from the experiment: a dataset, a task and scorers, run from the SDK, CLI or UI and compared side by side. Loop, its in-product agent, reads project data, writes scorers and runs evals, and since September 2026 Patterns and Topics group production behaviour for review. Many of these newer features are in public preview. Workflow evals through provider batch APIs, also new, are promoted as roughly half the cost of standard eval runs.
Opik covers the same loop with datasets, experiments, Test Suites written as plain-text assertions and resumable evaluations, and adds more than 30 built-in judge metrics, an Agent as a Judge that reads traces with tools, and a Prompt Optimizer with six algorithms. Ollie, its assistant, analyses traces and writes fixes with test cases. Online evaluation rules score sampled production traces, spans and threads, and custom Python metrics can run as online rules as well.
Both have human review. Braintrust supports assignment to reviewers, blind reviews and required scores, though its free Starter plan allows one human review score per project. Opik uses annotation queues, which since October 2026 can fill automatically from feedback-score conditions.
Braintrust documents no inline blocking, guardrails or approvals on live agent actions. Online scoring runs asynchronously after logging, and alerts to Slack, PagerDuty or webhooks fire after the fact. Loop asks for approval only for its own edits inside Braintrust.
Opik ships guardrails for PII, topics, prompt injection, judge checks and custom classifiers. They run as a blocking validate() call that fails closed and raises an exception, so your application must call them and handle the result. PII and topic checks are English only, and self-hosted guards need a separate guardrails server, with a GPU recommended. There is no approval or hold workflow for agent actions.
Each has a gateway with caveats. Braintrust's hosted gateway is free, in public preview and without an SLA. Opik's LLM Gateway is in beta and documented for development and testing, not production; Opik also traces through third-party gateways such as LiteLLM, Portkey and OpenRouter.
The two meter different things. As of 20 October 2026, Braintrust charges a platform fee plus processed data and scores: Starter is free with 1 GB, 10K scores and 14-day retention, and Pro is $249 a month with 5 GB, 50K scores and 30-day retention. Every Braintrust plan includes unlimited users. Opik charges by span: Free Cloud covers 25k spans a month for up to 10 team members, and Pro Cloud is $19 a month for 100k spans and up to 50 members, then $5 per 100k, with 60-day retention on both.
A large team may find Braintrust's unlimited users offset its higher base fee; a team with modest traffic will usually pay less on Opik, or nothing if it self-hosts. Check Braintrust's pricing and Opik's pricing against your own volumes.
We sell the layer this section describes. Read it with that in mind.
Stated plainly, since Prefactor publishes this page: Prefactor is an agent evaluation platform that scores each step of a run and can act on what it finds through runtime policies, approval holds and a kill switch scoped to a run, an agent or a team. Braintrust documents no run-time checks; Opik's guardrails block a call when your code invokes them, but it has no approval holds or tool-call policies. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if prompt iteration or free self-hosting is what you need, Braintrust or Opik is the better pick. See the full comparison of 15 platforms.
Reviewed against public product and documentation pages on October 20, 2026. If a vendor has changed a feature, product name, or positioning since then, send a correction and we will update it. Numbered source links in the page body point to the ordered sources below.
Prefactor scores each step of every agent run and can hold or stop an action when a score says it should not happen. Start on the free tier: 5,000 spans a month, no credit card required.
Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.