Galileo pricing

The Plans, as Published

The page also says annual billing on Pro "saves 33%", and at the listed rate Pro is a $1,200 commitment for the year. Only Free has a sign-up button: Pro, despite its listed price, routes to "Book a Demo", as Enterprise does. The self-serve path at Galileo is the Free plan; everything above it starts with a conversation.

Galileo is now named Splunk Agent Observability following Cisco's acquisition. The Galileo pricing page still publishes the plans above. If you are buying through Splunk or Cisco, confirm which price list and contract terms apply to the quote.

What Grows with Volume

The meter is traces. Free stops at 5,000 a month and Pro includes 50,000, and the Pro plan notes that pricing "scales based on number of traces" - but the per-trace rate above 50,000 is not on the page. A cost model beyond Pro's allowance is therefore a question for sales, not arithmetic.

Three things decide how fast you reach the limits:

• What counts as a trace. An agent session with twenty tool calls could be one trace or several, depending on how it is instrumented. Ask how Galileo counts a multi-step agent run before you estimate.
• How much you evaluate. Unlimited custom evals on Free means the number of evaluations is not what you pay for; traces are. That is generous for a team running many metrics on modest traffic.
• Whether you need runtime checks. Hosted real-time guardrails are Enterprise, and so is Luna-2, the model family Galileo uses to make scoring every trace cheap. The moment you need either, you are pricing Enterprise.

A worked example, assumptions stated: an agent that handles 1,500 conversations a day, one trace each, sends about 45,000 traces in a 30-day month. That fits inside Pro. At 3,000 conversations a day, or if each conversation produces two traces, it is about 90,000 traces - over Pro's allowance, at a rate the page does not publish.

What Only Enterprise Gets

Most of what makes Galileo distinctive at runtime is on the quoted tier:

• Real-time guardrails, listed only under Enterprise on the pricing page.
• Luna-2, Galileo's 3B and 8B evaluation models. Its docs say "Luna-2 is only available in the Enterprise tier" and list a rate of $0.02 per million tokens.
• Managed Agent Control. Galileo's release notes list the centrally managed version as available to enterprise customers only. The open-source Agent Control is free under Apache-2.0 on any plan, provided you run its server.
• SSO, VPC and on-prem, plus a dedicated CSM, low-latency dedicated inference servers and forward-deployed engineers.

Putting SSO on the top tier is common in this market; Raindrop does the same, and Future AGI sells SSO as a paid add-on. It matters more at Galileo because its own pitch is that "offline evals become production guardrails". On Free and Pro you get the evals; the hosted guardrails are the upgrade.

Against the Category's Published Prices

At the entry tier, Galileo Pro and Failproof AI Team have almost the same base price and a 50,000-session headline allowance, although Galileo meters traces while Failproof AI meters one run per agent session and counts evaluations separately. Galileo Free includes unlimited users and custom evaluations. Failproof AI Team includes 50,000 runs and 50,000 evaluations, publishes both overage rates, and adds automatic recurring-failure analysis, findings with recommended fixes, alerts and ownership. Scale publishes SSO/SAML at $599 a month. Full Failproof AI plans are on /pricing.

Future AGI is harder to put in the same table because it meters differently: a free tier with 50GB of storage, 2K AI credits and 100K gateway requests, pay-as-you-go above that, and add-ons at $250 a month (SOC 2, OAuth SSO, 90-day retention), $750 (SAML and SCIM, a HIPAA BAA, one-year retention) and $2,000 (ABAC, data masking, a dedicated engineer). Against that spread Galileo sits in the middle of the market: two plans with a listed price, and a third - the one with hosted guardrails, Luna-2 and on-prem deployment - that you negotiate.

What to Ask Before You Sign

• How is a trace counted for a multi-step agent, and what is the per-trace rate above Pro's 50,000?
• Is Pro available on monthly billing, and at what price?
• Which of the 20+ built-in metrics run on Free and Pro, and on which model?
• Does an Enterprise quote include Luna-2 usage, or is it metered at the listed $0.02 per million tokens on top?
• After the move to Splunk Agent Observability, which price list and contract terms apply at renewal?

Questions to ask an agent eval vendor has the longer checklist.

When the Free Plan Is Enough

For a team evaluating one agent on modest traffic, Galileo Free is a real product rather than a trial: 5,000 traces a month, unlimited users and unlimited custom evals cover prototyping and a small production agent. If you do not need runtime blocking, SSO or Luna-2, you may never need a quote. And if blocking is the only thing you need, the open-source Agent Control for agents you build yourself, or Failproof AI's free CLI for coding agents, costs nothing but the time to set it up.