Fiddler AI Control Plane
AI observability platform: evaluation, monitoring and real-time guardrails for ML models and generative AI/agentic applications in production; also detects and redacts PII, PHI and secrets before a prompt reaches the model and before a response reaches the developer.
Find alternatives to Fiddler AI Control Plane
What it does
- Evaluation and observability
Used for
- Guardrails and safety
Features 8
- Agent control plane viewSingle-pane view of the performance of all agents, with every action, token and dollar by user and team.
- Inline policy enforcementDetects and redacts PII, PHI and secrets in the agent request and response path.
- Auditable GovernanceEvery enforcement decision is recorded with who triggered it and what happened.
- Fiddler Centor ModelsModels that power guardrails and evaluations in-environment with no external LLM API calls.
- Evaluator RulesDefines how automated evaluations are applied to an application’s spans.
- User-defined metrics (FQL)Defines measurements that fit an agentic application’s requirements.
- AnnotationsAttaches human evaluation scores to individual spans.
- Fiddler Evals SDKRuns batch experiments with 13 pre-built evaluators or user-defined metrics.
How it compares
-
Arize AX
Fiddler vs. Arize -- Arize Alternatives
Official documentation · 19 Sep 2026
Sources
- Pricing
-
Three named tiers read live: Free $0 (real-time guardrails); Developer $0.002 per trace; Enterprise custom.
- Description
-
Official documentation · 9 Sep 2026
Own About page: evaluation, monitoring and guardrails for ML and generative AI/agentic applications in production.
- Used for
-
Official documentation · 24 Oct 2026
'Detect hallucinations, toxicity, bias, PII/PHI leakage, jailbreaks, and policy violations in real-time.' (https://www.fiddler.ai/ai-observability)
- Constraints
-
Derived from the security facts on this page, from the same source.
- Security
-
Platform-wide: fiddler.ai/security describes the Fiddler platform; fiddler-auditor (open-source library) left out. HIPAA page says 'in compliance' without a BAA, so hipaa-baa not recorded. SSO named without SAML, so sso-saml not recorded. Training, ZDR, residency not stated. Trust Center (trust-center.fiddler.ai) not read.
- Sold within
-
Bought and metered on its own pricing page -- no parent offering it ships inside.
- URL
-
Official documentation · 9 Sep 2026
Fetched 2026-09-09: markets as an independent AI observability product.
- Deployment
-
Enterprise tier names "flexible deployment options (SaaS, VPC, or on-premise)" explicitly -- mapped to this site's vocabulary as saas / private-cloud / self-hosted.
- Interfaces
-
Official documentation · 14 Oct 2026
Page: 'Python Client SDK - Core SDK client for ML/LLM monitoring' / 'REST API - Complete HTTP API for all platform features'.
- Intended for
-
Official documentation · 13 Oct 2026
Page: 'Fiddler is built for enterprises that need one platform to observe, evaluate, and enforce policy across every agent'.
Also from Fiddler AI
-
Fiddler Auditor Developer tool
Source-available library for red-teaming and robustness testing of LLMs, run independently of the hosted Fiddler platform. Fiddler's own page calls it "open-source"; its repository's actual licence is the Elastic License 2.0, which is not an OSI-approved open-source licence (it restricts offering the software as a competing hosted service), so it is recorded here by its real terms rather than the vendor's label.
-
Fiddler Agentic Observability Application
Tracing and monitoring module for multi-step agent workflows within the Fiddler AI Control Plane, tracking tool calls, reasoning steps and failures across an agent's execution; also blocks risky tool use and data leaks.
-
Governance module within the Fiddler AI Control Plane providing model inventories, policy enforcement and audit-ready reporting for regulatory compliance (e.g. EU AI Act) across an organization's AI systems.
-
Fiddler Continuous Evaluations Application
Ongoing automated evaluation module within the Fiddler AI Control Plane that re-scores production LLM/agent outputs against quality and safety metrics over time, not just at launch.
-
Fiddler Guardrails Application
Runtime safety layer within the Fiddler AI Control Plane that screens LLM/agent inputs and outputs for hallucination, toxicity, PII and jailbreak attempts before they reach users.
Open Fiddler AI in the directory
Something wrong here? Send a correction — quote this product id: fiddler-ai-platform.