Fiddler AI
Platform from Palo Alto for monitoring, evaluating and safeguarding AI agents, LLM applications and classic machine learning models.
What it does
Fiddler records what AI agents and LLM applications do in production: every step, the tokens used and the cost, broken down by user and team. Evaluations check responses from testing through to production, optionally with your own evaluators. Guardrails sit directly in the request path, detect and redact personal data, health data and secrets, block prompt injection and jailbreaks, and flag hallucinations and toxic content. For classic predictive models there is monitoring with more than 100 predefined metrics. You connect it via SDKs for LangGraph, LangChain and Strands or via OpenTelemetry.
Who it suits
Data science and platform teams in larger companies that run several AI applications in production and need to document quality, cost and risk in a traceable way, for example in regulated industries.
Data protection for businesses
Fiddler Labs is based in the USA. The vendor does not mention EU data hosting; in the Enterprise plan, however, you can run the platform in your own VPC or on-premises so the data stays in your environment. The vendor does not mention a data processing agreement on its public pages, nor whether customer data is used for training – clarify both with sales before deployment. The vendor states SOC 2 Type II and HIPAA compliance.
Limits
The free plan covers only the guardrails. In the Developer plan, costs grow with the number of traces. Self-hosting is only available in the Enterprise plan. Connecting your applications requires programming skills. On 3 November 2026, CrowdStrike announced that it will acquire Fiddler and integrate it into its Falcon platform; the acquisition has not yet closed, and the vendor does not say how plans and contracts will change afterwards.