Galileo vs LangSmith

Galileo is an evaluation, observability and guardrails platform for generative AI applications and AI agents. It has been owned by Cisco since May 2026: Cisco announced its intent to acquire Galileo Technologies, Inc. on April 9, 2026, and said on May 22, 2026 that the acquisition was complete. Terms weren't disclosed. Galileo is now part of Splunk, a Cisco company, and its documentation says that as of August 7, 2026 "Galileo is now Splunk Agent Observability." Customers who onboarded before that date use Galileo's documentation; later customers use Splunk's. galileo.ai's pricing page now redirects to Splunk's pricing page. Galileo was founded in 2021 by Vikram Chatterji (CEO), formerly of Google AI, Atindriyo Sanyal, formerly of Apple and Uber AI, and Yash Sheth, formerly of Google, and was based in San Francisco. Before the acquisition it raised $68 million, most recently a $45 million Series B led by Scale Venture Partners in October 2024; it then said six Fortune 50 companies, including Comcast and Twilio, were customers. The platform logs traces and spans from agent frameworks (LangChain and LangGraph, CrewAI, the OpenAI Agents SDK, Strands Agents, the Vercel AI SDK, and any OpenTelemetry or OpenInference source) and scores them with metrics such as context adherence, correctness, tool selection quality, action completion, toxicity, PII, prompt injection and sexism/bias. Many metrics run on Luna-2, Galileo's small evaluation models, which require an enterprise plan; Galileo says they cut production monitoring costs by up to 97% with latency under 200 ms. Signals (January 2026) detects failure patterns across production traces and can turn them into custom metrics. Agent Control, an open-source (Apache 2.0) control plane released in March 2026, applies central policies to agents (deny, steer, warn, log or allow) using Luna, NVIDIA NeMo, Amazon Bedrock or custom evaluators. It replaced Galileo's older Protect guardrails, which were deprecated in June 2026. Under Splunk, Galileo's technology is being added to the AI Agent Monitoring feature of Splunk Observability. In June 2026 Splunk described Splunk Agent Observability as a new on-premises offering powered by Galileo; in October 2026 it said the product had started rolling out in September 2026 on premises, in Cisco Cloud Control and in Splunk Observability Cloud, and its pricing page lists it for Observability Cloud and self-hosted deployments. **AI compliance coverage** Galileo doesn't document support for the EU AI Act, the NIST AI RMF or ISO 42001. Its trust page says it has a SOC 2 Type II report, provides HIPAA-compliant infrastructure and can sign business associate agreements (BAAs). What it offers a compliance program is evidence of how models and agents behave (evaluation scores, traces and guardrail decisions), not framework controls or an AI inventory. **Who it's for** Engineering teams building LLM applications and agents that want evaluation, production monitoring and runtime guardrails in one tool. New customers now buy through Splunk, so it fits Splunk Observability customers most naturally.

LangSmith is an LLM observability platform that provides tracing, monitoring, and evaluation for AI agents and LLM applications. It offers native tracing for agent frameworks, cost and latency tracking, online LLM-as-judge evals, custom dashboards, and alerts via webhooks or PagerDuty. Framework-agnostic with SDKs for Python, TypeScript, Go, Java, and OpenTelemetry support, it works with OpenAI, Anthropic, LlamaIndex, and custom stacks. Typical buyers are engineering teams building production LLM apps needing visibility into agent behavior, debugging failures, and performance optimization. Enterprise plans include self-hosted and BYOC options for data residency.LangSmith homepage Pricing