Both trace agents, run experiments and score production traffic. Galileo, now Splunk Agent Observability, has the stronger run-time guardrails and fits Splunk estates; Opik is open source, free to self-host and self-serve from $19 a month. How you buy software often decides this before features do.[3][4]
The difference between Galileo and Opik starts with how you buy them: Galileo is now sold as Splunk Agent Observability after Cisco completed its acquisition in May 2026, with annual pricing and no free tier listed, while Opik, from Comet, is an Apache-2.0 platform with a free self-hosted version and a $19-a-month cloud tier. Galileo's Agent Control can block harmful model and tool inputs and outputs during a run without agent code changes; Opik's guardrails are called from your code.
| Decision factor | Galileo | Opik |
|---|---|---|
| Ownership | Acquired by Cisco (completed May 2026); sold as Splunk Agent Observability since 7 August 2026 | Comet ML, independent; reports $70M in total funding |
| Runtime guardrails | Yes: Agent Control checks model and tool inputs and outputs during a run | Partial: guardrails your code calls; they fail closed |
| Licence | Proprietary platform; SDKs and Agent Control are Apache-2.0 | Apache-2.0 for the core platform and SDKs |
| Self-hosting | On-premises, including air-gapped (commercial) | Free; all features except user administration |
| Prompt tooling | Prompts and playground; version control not documented | Prompt Library with environment-tagged versions, plus an optimiser |
| Free tier | None listed for Agent Observability | 25k spans a month, 60-day retention, up to 10 members |
| Paid pricing | From $100 a month per 1.2 million spans a year, billed annually | Pro Cloud: $19 a month for 100k spans, then $5 per 100k |
| Capability | Galileo | 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 | Partial | ✓ | — |
| 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 |
Cisco announced its intent to acquire Galileo on 9 April 2026, and an update dated 22 May 2026 says the acquisition has completed. Since 7 August 2026 the product has been sold as Splunk Agent Observability and is being folded into Splunk Observability Cloud. Customers who joined before that date still use Galileo's documentation; later customers use Splunk's. The old pricing page redirects to Splunk, which lists annual billing and no free tier for this product.
Opik is built by Comet ML, which reports $70M in total funding and has no ownership change on record as of 20 October 2026. It is self-serve: you can start on the free cloud tier, pay monthly from $19, or self-host the Apache-2.0 release. If you do not already use Splunk, that difference in buying process may decide the question before any feature comparison.
Galileo's Agent Control evaluates model and tool inputs and outputs while an agent runs and can block prompt injection, personal data leaks and harmful content using rules set in one place, without changes to agent code. It replaced the older Protect product in June 2026, and its core was released as open source under Apache-2.0 in March 2026; the hosted version is an enterprise feature. Blocking content is documented; approval workflows for actions are not.
Opik's guardrails cover PII, topics, prompt injection, judge checks and custom classifiers. They run as a blocking validate() call that fails closed and raises an exception, which means your code decides where checks happen and must handle the result. PII and topic checks are English only, and self-hosted guards need a separate guardrails server, with a GPU recommended. For a team that wants central rules covering tool inputs and outputs without touching agent code, Agent Control covers more.
Galileo offers preset and custom LLM judges, Autotune to adjust judges from feedback, Luna-2 small models as lower-cost evaluators and Python code metrics. Multi-turn experiments arrived in June 2026 and Annotation Queues became generally available in August. One limit to check: the Splunk-hosted edition supports five preset Luna evaluators and no custom code evaluators or Luna fine-tuning, which are available on-premises only. Its prompts page documents storage, templating and a playground but not version control.
Opik offers more than 30 built-in metrics, custom judge prompts, an Agent as a Judge and custom Python metrics that run in experiments and as online rules on production threads, on its cloud and self-hosted versions alike. Its Prompt Library keeps environment-tagged versions, and the Prompt Optimizer runs six optimisation algorithms. Since July 2026 you can cap spend per evaluation for judges.
Instrumentation is comparable. Galileo accepts OTLP traces, supports OpenInference instrumentors in Python, and has native integrations for the OpenAI Agents SDK, LangGraph, CrewAI, Google ADK, Strands, Microsoft Agent Framework and A2A multi-agent tracing. Its SDKs are Python and TypeScript. Opik accepts OTLP over HTTP only, lists most of the same frameworks, and adds SDKs for Java and .NET.
As of 20 October 2026, Splunk lists Agent Observability from $100 a month per 1.2 million spans a year, billed annually, including 500 million Luna tokens and five Luna fine-tuning runs a year. No free tier is listed for this product. It runs as SaaS in Splunk Observability Cloud or on-premises, including air-gapped, as a commercial deployment.
Opik's Free Cloud covers 25k spans a month for up to 10 team members with 60-day retention, and Pro Cloud is $19 a month for 100k spans and up to 50 members, then $5 per 100k. Self-hosting is free, and Enterprise adds on-premises or fully hosted options with SSO. Check Splunk's pricing for Galileo and Opik's pricing against your own span 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 acts on what it finds through runtime policies on tool calls and data access, approval holds for risky actions, and a kill switch. Of these two, Galileo's Agent Control is the closest to that, though it documents content blocking rather than approval holds; Opik's guardrails depend on calls in your code. Prefactor is self-serve with a free tier, but it has no prompt playground or prompt registry and its platform is hosted. 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.