Fiddler AI

Fiddler AI positions itself as the control plane, or system of trust, for first-party and third-party AI agents across the lifecycle. Its offerings include Continuous Evaluations, Agentic Observability, Policy Enforcement (guardrails), ML Observability, and AI Governance, Risk Management, and Compliance. The platform is supported by Fiddler Centor Models, which the page says run evaluations and real-time policy enforcement in-environment without external LLM API calls. The page states deployment options include SaaS, VPC, and AWS GovCloud, and lists partnerships with Amazon SageMaker AI, Google Cloud, NVIDIA NIM and NeMo Guardrails, Databricks, and Datadog. Solutions are highlighted for government, financial services, healthcare, and insurance, and the page features case studies with Nielsen and other organizations.

Key features

• Continuous evaluations, including in-environment evaluation from testing to production
• Agentic observability with visibility into actions, tokens, and cost by user and team
• Inline policy enforcement on the agent request and response path that detects and redacts PII, PHI, and secrets
• Fiddler Centor Models for guardrails and evaluations without external LLM API calls
• ML observability for predictive models
• Auditable governance that records enforcement decisions with who triggered them and what happened
• 100+ pre-defined metrics plus custom metrics
• Deployment in SaaS, VPC, or AWS GovCloud
• Model- and cloud-agnostic support, including Azure OpenAI, Amazon Bedrock, and LangGraph per the FAQ
• TCO calculator for evaluation costs

Best suited for

• Enterprises deploying AI agents in regulated industries such as financial services, healthcare, insurance, and government
• Software and AI engineers, data scientists, platform and DevOps teams, and security and compliance teams
• Organizations needing centralized governance and audit evidence for AI systems
• Teams seeking to reduce evaluation costs tied to external LLM API calls

Pricing

Three plans are offered: a Free plan with real-time guardrails, a Developer plan at $0.002 per trace (usage-based), and an Enterprise plan with custom pricing. Full details are on the pricing page.