Production AI-agent teams
Observe agent workflows end to end and investigate failures across prompts, models, tools, retrieval steps, and infrastructure.
Independent tool overview
Galileo (now Splunk Agent Observability) at a glance
Galileo's AI evaluation and observability technology now powers Splunk Agent Observability, a platform for tracing, evaluating, securing, and controlling the cost of production AI agents.
Visit the official Galileo (now Splunk Agent Observability) site ↗
- Current name
- Splunk Agent Observability
- Best for
- Engineering teams running AI agents in production
- Core jobs
- Tracing, evaluation, tokenomics, and guardrails
- Current pricing
- Contact Splunk for a quote
- Legacy pricing
- Free and $100/month Pro plans remain listed for the legacy service
- Acquisition completed
- May 22, 2026
Overview
What Galileo (now Splunk Agent Observability) is
Galileo began as a specialist platform for evaluating and observing generative AI applications. Cisco completed its acquisition of Galileo in May 2026, and the product became Splunk Agent Observability for new customers in August 2026. The current product brings Galileo's evaluation models and guardrails into Splunk's broader application and infrastructure observability stack.
The platform is designed for teams operating AI agents and LLM applications in production. It traces multi-step workflows, scores output quality and safety, tracks token usage and cost, and can enforce runtime guardrails. Its central advantage is connecting an agent's prompts, model calls, tools, retrieval steps, and infrastructure so teams can investigate both behavioral and system failures in one place.
Galileo's legacy workspace and documentation remain relevant to customers who joined before the transition. New buyers should evaluate the Splunk product and commercial terms rather than assume the legacy Galileo plans or interface are still the onboarding path.
Use cases
Who Galileo (now Splunk Agent Observability) is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
AI quality and evaluation programs
Run built-in or custom evaluators across experiments and production traffic instead of relying only on manual spot checks.
Security and governance teams
Detect or block risks such as prompt injection, sensitive-data leakage, unsafe tool use, hallucinations, and biased outputs.
Teams controlling inference costs
Attribute token and model costs to agents, requests, and workflows, then compare spend with quality and business outcomes.
Capabilities
Core Galileo (now Splunk Agent Observability) features
Agent workflow tracing
Maps requests through model calls, tools, retrieval steps, and downstream infrastructure so teams can isolate where a workflow failed.
Online and offline evaluations
Scores correctness, relevance, tool use, safety, and custom criteria during experiments and in production.
Luna evaluation models
Purpose-built small language models provide lower-latency, lower-cost evaluations and guardrails than repeatedly using frontier models as judges.
Tokenomics
Tracks token use and model cost by request, agent, and workflow alongside latency, errors, and quality signals.
Runtime guardrails
Can block or steer risky interactions involving prompt injection, unsafe tool actions, and PII, PHI, or PCI leakage.
Framework integrations
Legacy Galileo documentation supports SDKs, APIs, OpenTelemetry, OpenAI, LangChain and LangGraph, CrewAI, Google ADK, Vercel AI SDK, and other agent frameworks.
Process
How the Galileo (now Splunk Agent Observability) workflow works
Step 1
Instrument the application
Connect an SDK, supported framework integration, API, or OpenTelemetry tracing to capture agent activity.
Step 2
Define quality signals
Choose built-in metrics and add custom evaluators that reflect the application's task, risks, and success criteria.
Step 3
Test before release
Run datasets and experiments to compare prompts, models, agent versions, and workflows before deployment.
Step 4
Monitor production
Review traces, quality scores, latency, errors, token cost, and infrastructure signals across live traffic.
Step 5
Protect and improve
Set alerts or guardrails, investigate root causes, and feed production failures back into evaluation datasets.
Cost
Galileo (now Splunk Agent Observability) pricing and free plan
Galileo became Splunk Agent Observability for new customers on August 7, 2026. Splunk does not publish a standalone price for the current product, so new buyers should request a quote. The legacy Galileo pricing page still lists Free and Pro plans for customers on the earlier service; those figures should not be treated as guaranteed Splunk pricing.
Splunk Agent Observability
Custom quote
The current onboarding path for new customers.
- Contact Splunk sales
- Evaluation, observability, tokenomics, and guardrails
- Commercial terms depend on deployment and scale
Legacy Galileo Free
$0/month
The public legacy plan for developers and small teams.
- 5,000 traces per month
- Unlimited users
- Unlimited custom evaluations
Legacy Galileo Pro
$100/month billed yearly
The legacy paid starting point, with pricing that scales by trace volume.
- 50,000 traces per month at the listed starting price
- Standard role-based access control
- Advanced analytics and Slack support
Pricing checked . Check current pricing at the source ↗
Assessment
Galileo (now Splunk Agent Observability) strengths and limitations
Where it stands out
- Combines agent behavior, model quality, cost, security, and infrastructure context in one investigation workflow.
- Supports experiments before release and continuous evaluation after deployment.
- Purpose-built Luna models make broad production evaluation more practical than sampling every interaction with a frontier judge.
- Runtime guardrails can act on risks rather than merely reporting them later.
- Works across major model providers and popular agent frameworks instead of locking teams to one LLM stack.
What to consider
- The Galileo-to-Splunk transition creates two sets of documentation, branding, onboarding paths, and pricing references.
- Current Splunk Agent Observability pricing is not publicly itemized, which makes early cost comparison difficult.
- Useful evaluations still require representative datasets, domain-specific criteria, and periodic human calibration.
- Full-stack value is greatest for teams already using or willing to adopt Splunk; smaller teams may prefer a lighter standalone tool.
- Tracing production prompts and outputs requires careful access controls, redaction, retention policies, and consent practices.
Compare
Galileo (now Splunk Agent Observability) alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Coding
AgentOps
A focused agent monitoring and debugging platform that may be easier for smaller development teams to adopt.
Explore AgentOps →Coding
Helicone
An LLM observability gateway with usage, cost, latency, prompt, and evaluation workflows.
Explore Helicone →Questions
Galileo (now Splunk Agent Observability) FAQs
What happened to Galileo AI?
Cisco completed its acquisition of Galileo Technologies on May 22, 2026. Galileo's technology now powers Splunk Agent Observability, and the legacy documentation says new customers after August 7, 2026 should use the Splunk product documentation.
Is Galileo still active?
Yes, but as part of Splunk Agent Observability. Existing Galileo customers may still use the legacy workspace and documentation, while new customers are directed to Splunk.
What does Splunk Agent Observability do?
It traces AI-agent workflows, evaluates output quality and safety, tracks token usage and cost, correlates agent behavior with infrastructure, and applies runtime guardrails.
How much does Galileo cost?
New customers need a custom quote from Splunk. The legacy Galileo page lists a free plan with 5,000 monthly traces and a Pro plan starting at $100 per month when billed yearly, but those are not guaranteed to be the current Splunk terms.
Can Galileo evaluate all production traffic?
Splunk positions its purpose-built Luna models as economical enough to evaluate and guardrail 100% of production interactions. Actual coverage and cost depend on the selected deployment and contract.
Does it replace application monitoring?
No. Agent observability adds model, prompt, tool, retrieval, quality, safety, and token-cost context. Splunk's differentiator is correlating those signals with application and infrastructure telemetry.
Bottom line
Our Galileo (now Splunk Agent Observability) verdict
Galileo remains a serious option for enterprise AI-agent reliability, now inside Splunk. Its strongest case is for teams that need to connect agent behavior, evaluation, security, cost, and infrastructure at production scale. Smaller teams should compare the custom-priced Splunk path with lighter standalone observability tools, and legacy Galileo customers should confirm how their account will migrate.
Visit Galileo (now Splunk Agent Observability) website ↗
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.