Fiddler AI

What it does

Fiddler AI is a platform for monitoring, evaluating, and enforcing policies on AI agents and LLM applications. It takes in agent prompts, responses, traces, and spans from pre-production testing or live traffic, and produces evaluation scores, observability views, guardrail decisions, and audit records of enforcement actions.

Teams evaluate agents against curated datasets before deployment, then observe live sessions in Trace Explorer and feed annotated failures back into test suites. Guardrails check prompts and responses for hallucinations, jailbreaks, PII, and unsafe content, with a stated latency under 80ms. Fiddler states that evaluations run in the customer's environment through its Fiddler Centor Models, without external LLM API calls.

Fiddler fits enterprise and developer teams building agentic or LLM applications that need runtime controls and governance. Access is through a demo request for the Free and Developer plans and a contact-sales route for Enterprise, with deployment options of SaaS, VPC, or on-premise.

What it can do

• Real-time guardrails that evaluate prompts and responses for hallucinations, jailbreaks, PII/PHI, and unsafe content
• Pre-production and production evaluations with 80+ out-of-the-box metrics and custom evaluators
• Trace Explorer for filtering, sorting, and searching agent spans and sessions
• Enterprise-grade governance with audit records of enforcement decisions
• Unified observability across agentic and predictive ML systems

Works with

• NVIDIA NeMo Guardrails
• Amazon SageMaker Studio

Before you choose

• Pricing for the Developer plan is per trace, so evaluation volume drives cost; the page does not state monthly or annual billing for it.
• Enterprise deployment options (VPC or on-premise) and customized onboarding are handled through sales.
• Evaluation coverage and guardrail behavior depend on configured thresholds and policies.

Listed under

• Coding
• Agent evaluation
• AI agents
• AI governance
• AI observability
• Compliance
• Enterprise

Get started

Enterprise access starts with a demo request; developer documentation is available.

1. Choose the safety, security, privacy, and accuracy policies your agents need to follow, including thresholds.
2. Write three to five lines of code to initialize your HTTP client.
3. Call the Fiddler Guardrails API using pre-built code examples in NodeJS, Python, cURL, or any HTTP library.

Developer is listed at $0.002 per trace; Enterprise pricing is arranged through sales. Confirm billing cadence and any Free plan usage limits with Fiddler before purchase.

Source: fiddler.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Fiddler AI: The Control Plane for Enterprise AI Agents ↗www.fiddler.ai
• AI Agent Evaluations Across the Lifecycle | Fiddler AI ↗www.fiddler.ai
• Fiddler Guardrails: Real-Time Policy Enforcement for Agents | Fiddler AI ↗www.fiddler.ai
• Plans and Pricing | Fiddler AI ↗www.fiddler.ai
• Introduction to Fiddler - Fiddler Documentation ↗docs.fiddler.ai

Company website

fiddler.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.