Both trace agents, run experiments and score production traffic. Galileo, now Splunk Agent Observability, suits Splunk estates and teams that need checks on tool inputs and outputs during a run; Logfire is the self-serve option for teams that want agent traces inside whole-application OpenTelemetry tracing. Your observability vendor and how you buy usually decide it.[3][4]
Galileo, sold as Splunk Agent Observability since 7 August 2026 after Cisco completed its acquisition in May 2026, is an evaluation platform with run-time guardrails on model and tool inputs and outputs; Logfire is Pydantic's OpenTelemetry observability platform with evals, prompt versioning and an AI gateway. Galileo's Agent Control can block content during a run, while Logfire's gateway checks model calls only. Logfire has self-serve plans starting free; Splunk lists annual, paid pricing with no free tier.
| Decision factor | Galileo | Logfire |
|---|---|---|
| Ownership | Cisco (completed May 2026); sold as Splunk Agent Observability since 7 August 2026 | Pydantic, independent |
| Run-time checks | Agent Control blocks harmful model and tool inputs and outputs | Gateway guardrails and spend caps on model calls |
| Evaluators | LLM judges, Luna-2 small models; custom code metrics on-premises only | pydantic-evals LLM judges and code scorers |
| Prompt tooling | Prompts and playground; version control not documented | Prompt versioning with labels and rollout targeting |
| Self-hosting | On-premises, including air-gapped (commercial) | Enterprise, on Kubernetes via Helm |
| Free tier | None listed for Agent Observability | Personal: 10M records a month, 1 seat |
| Paid pricing | From $100 a month per 1.2 million spans a year, billed annually | Team: $49 a month; Growth: $249 a month |
| Capability | Galileo | Logfire | 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 | ✓ | Partial | ✓ |
| Prompt versioning and playground | Partial | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | ✓ | Partial | ✓ |
| Gateway or proxy in the request path | — | ✓ | — |
| Licence and deployment | |||
| Open source | Partial | Partial | Partial |
| Self-hosting | ✓ | Partial | Partial |
Cisco announced its intent to acquire Galileo in April 2026 and completed the deal in May. Since 7 August 2026 the product has been branded Splunk Agent Observability and is being integrated into Splunk Observability Cloud. Two documentation sites now exist, split by onboarding date: customers who joined before 7 August use Galileo's, later customers use Splunk's. Galileo's pricing page redirects to Splunk.
In practice, buying Galileo today means buying a Splunk product on annual terms. That fits organisations that already run Splunk, and is a slower start for a small team that wants to trial a tool first. Pydantic, which makes Logfire, is independent, and Logfire offers a self-serve free plan. Its SDK releases have continued through October 2026, including OpenTelemetry 1.45 support on 13 October.
This is the clearest functional gap between the two. 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 centrally, without changes to agent code. Its core was released as open source under Apache-2.0 in March 2026 and can be self-hosted from that repository. It replaced Galileo Protect, deprecated in June 2026, so Protect users need to move. Blocking content is documented; approval workflows for actions are not.
Logfire's checks live in its AI gateway, part of Logfire since March 2026. Per-rule guardrails can observe, flag, redact or block model calls containing secrets, personal data or policy violations, and spend caps stop requests once a budget is used. They apply only to model traffic routed through the gateway, so a tool call the agent makes directly is not inspected.
Instrumentation is comparable. Galileo has Python and TypeScript SDKs, accepts OTLP traces and supports OpenInference instrumentors in Python, with native integrations for the OpenAI Agents SDK, LangGraph, CrewAI, Google ADK, Strands and Microsoft Agent Framework, plus A2A multi-agent tracing since May 2026. Logfire covers many of the same frameworks and adds the HTTP, database, log and metric data around them.
Galileo runs experiments on datasets with preset and custom metrics, multi-turn experiments since June 2026, and Autotune to adjust LLM judges from feedback. Luna-2 small models act as lower-cost judges. Annotation queues became generally available in August 2026. One limit to check: the Splunk Observability Cloud edition supports five preset Luna evaluators and no custom code evaluators, which are available on-premises only.
Logfire's evals use pydantic-evals, so LLM judges and plain-code scorers run the same way in every deployment. Live evaluations run in your application at a sample rate and report as OpenTelemetry events. Its human review is less mature: run annotations are in beta and queues in early access. On prompts the order reverses. Galileo stores, templates and reuses prompts with a playground but does not document version control; Logfire has immutable prompt versions, labels with rollout targeting, runtime fetch and a Prompt Playground that replays prompts from traces.
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. Logfire Personal is free for 10M records and one seat, Team is $49 a month for five seats, Growth is $249 a month with unlimited seats, and extra records cost $2 per million.
Both can run on your own infrastructure under a commercial agreement: Galileo on-premises including air-gapped sites, and Logfire Enterprise on Kubernetes via Helm or as a dedicated single-tenant deployment in any GCP region. Check Splunk's pricing and Logfire's pricing before you decide.
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. Of these two, Galileo's Agent Control is closer to that, since it checks tool inputs and outputs; Logfire's gateway covers model calls. Neither documents approval holds. 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.