Both trace agents, run experiments on datasets and score production traffic. Braintrust is the self-serve choice for teams iterating on prompts and scorers; Galileo, now sold through Splunk, is the one that can block harmful tool inputs and outputs while an agent runs. Whether you need checks inside the run usually decides it.[1][6]
Braintrust is an independent, self-serve evaluation platform built around experiments and prompt iteration, while Galileo is now Splunk Agent Observability and adds guardrails that act during a run. Cisco completed its acquisition of Galileo in May 2026, and the product has carried the Splunk name since 7 August 2026. Braintrust has a free tier and a $249 a month Pro plan; Splunk lists annual, span-based pricing and no free tier.
| Decision factor | Braintrust | Galileo |
|---|---|---|
| Ownership | Independent; $80M Series B in February 2026 | Cisco, completed May 2026; sold as Splunk Agent Observability since 7 August 2026 |
| Centre of gravity | Experiments: dataset, task, scorers, compare, with the Loop analysis agent | Evaluation plus run-time guardrails, inside Splunk Observability Cloud |
| Run-time guardrails | None; scoring and alerts run after logging | Yes: Agent Control checks model and tool inputs and outputs during a run |
| Prompt tooling | Versioned prompts, environments, playgrounds | Prompts and a playground; version control not documented |
| Gateway | Hosted gateway in public preview, no SLA | None |
| Self-hosting | Enterprise only, data plane only; Braintrust runs the control plane | On-premises, including air-gapped (commercial) |
| Free tier | Starter: 1 GB processed data, 10K scores, 14-day retention | None listed for Agent Observability |
| Paid pricing | Pro: $249 a month, unlimited users | From $100 a month per 1.2 million spans a year, billed annually |
| Capability | Braintrust | Galileo | 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 | ✓ | Partial | — |
| Act on findings | |||
| Runtime controls on live agent actions | — | ✓ | ✓ |
| Gateway or proxy in the request path | ✓ | — | — |
| Licence and deployment | |||
| Open source | Partial | Partial | Partial |
| Self-hosting | Partial | ✓ | Partial |
Cisco completed its acquisition of Galileo in May 2026, and since 7 August 2026 the product has been sold as Splunk Agent Observability inside Splunk Observability Cloud. Customers who joined before that date use Galileo's documentation; later customers use Splunk's. Galileo's pricing page now redirects to Splunk.
Braintrust remains independent and raised an $80M Series B in February 2026. It is bought self-serve, with a free tier and a published Pro price. The buying motion is the first practical difference: Braintrust can be started on its free plan without a sales conversation, while Splunk Agent Observability is an annual purchase that suits organisations already in the Splunk estate.
Deployment follows the same split. Braintrust runs as SaaS in US and EU regions; on Enterprise you can host its data plane in your own AWS, GCP or Azure account, but Braintrust always operates the control plane, including the UI and authentication. Splunk Agent Observability runs as SaaS in Splunk Observability Cloud or fully on-premises, including air-gapped, as a commercial deployment. For an organisation that must keep every component inside its own network, Galileo is the one of the two that documents it.
Galileo's Agent Control evaluates model and tool inputs and outputs during agent execution and can block prompt injection, personal data leaks and harmful content using rules set centrally, without agent code changes. It replaced the older Protect product in June 2026, and its core is open source under Apache-2.0. The documentation covers blocking content; it does not describe approval workflows for actions.
Braintrust does not document inline blocking or guardrails. Its online scoring rules run scorers asynchronously on production logs, and alerts, Patterns and behaviour specs judge traces after the fact. Loop asks for approval only for its own edits inside Braintrust. If your agents can do things with real consequences, Galileo covers more of the run itself; if you mainly need to measure and improve, the after-the-fact model is enough.
Braintrust is organised around the experiment. You bring a dataset, a task and scorers, run variants and compare them, then promote versioned prompts through environments and call them from code without redeploying. Loop reads results, writes scorers and runs evals, and since September 2026 Patterns and Topics group production behaviour, several of them in public preview. A gateway to OpenAI, Anthropic, Google and AWS models is also in public preview.
Galileo offers experiments, multi-turn evaluation since June 2026, annotation queues and Luna-2 small models as lower-cost judges. Its prompts page covers templating and reuse but does not document version control. One limit to check: the Splunk-hosted edition supports five preset Luna evaluators and no custom code evaluators, which are available on-premises only.
For teams whose daily work is prompt and scorer iteration, Braintrust has the deeper tooling. For teams who need evaluation next to infrastructure telemetry, Galileo's place in Splunk matters more.
As of 20 October 2026, Braintrust Starter is free with 1 GB of processed data, 10K scores and 14-day retention. Pro is $249 a month with 5 GB, 50K scores and 30-day retention, and every plan includes unlimited users. Enterprise is custom and is the only plan with self-hosting, which covers the data plane only. Splunk lists Agent Observability from $100 a month per 1.2 million spans a year, billed annually, including a Luna token allowance, with no free tier listed.
Retention is another difference. Braintrust keeps 14 days on Starter and 30 days on Pro by default, extendable to 180 days on Pro at $0.50 per GB a month. Splunk mentions extended data retention and extra Luna tokens as options without listing prices.
Braintrust meters data volume and scores; Splunk meters spans on an annual commitment. Check Braintrust's pricing and Splunk's pricing against your own traffic.
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 acts on what it finds through runtime policies on tool calls and data access, approval holds for risky actions, and a kill switch for a run, an agent or a team. Galileo's Agent Control is the closer of these two, since it blocks content during a run; Braintrust scores after logging. Neither documents approval holds. Prefactor has no prompt playground or prompt registry and is a hosted platform, so for prompt iteration Braintrust 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.