The Fiddler Control Plane for AI Agents

Agents are first-class enterprise actors, making autonomous decisions across your operations: reading files, calling tools, and pulling context faster than any reviewer can keep up with. However, most organizations don't fully know what their agents are accessing, spending, or exposing.

Fiddler is the system of trust for your agents. The Fiddler AI Control Plane gives you visibility, context, and control across the AI lifecycle, covering first-party, third-party, and coding agents. 

Visibility. Context. Control.

The Fiddler AI Control Plane helps you understand every agent decision, prevent unsafe actions, and govern agents at scale through four core capabilities. 

1. Reliable Evaluation
2. Continuous Monitoring
3. Enforceable Policy
4. Auditable Governance

One Control Plane. From Code to Production.

Fiddler governs agents where code is written and where they run in production. Deploy directly through Fiddler or through the gateway you already run, with no new infrastructure, agent-side integration, or SDK to embed. 

One Control Plane for Every Agent

Reliable Evaluation

Know how every agent performs, from pre-production to production.

• 100+ out-of-the-box and custom metrics for domain-specific use cases
• Evaluates agent performance, behavior, and reasoning chains
• Purpose-built Centor Models for low-latency evaluations, plus BYOJ for your preferred foundation models
  

Continuous Monitoring

See how every agent behaves and spends.

• Traces prompts, tool calls, decisions, and outcomes across first-party, third-party, and coding agents
• Captures inputs, outputs, actions, and execution context
• Fleet-wide cost and adoption intelligence, from individual developer to org-wide rollup

Inline Policy Enforcement

Stop unsafe actions before impact, not after.

• Guardrails for hallucination, safety, and jailbreaks in under 80ms
• Bidirectional PII/PHI and secrets detection and redaction on every request and response
• Enforces at the creation layer (IDE, CLI, MCP boundary) and at runtime

Auditable Governance

Centralized oversight and accountability across your agents.

• Agent registry: one system of record for every first-party, third-party, and coding agent
• Unified executive dashboard connecting agent behavior, performance, and cost to business KPIs
• Audit trails of decisions, actions, and policy outcomes, aligned with NIST and the EU AI Act

Works With The AI Stack You Already Have 

• ‍Gateways: 
  
  • ‍LiteLLM
  • AgentGateway
  • Kong Gateway‍
• Frameworks:‍
  
  • LangGraph
  • Amazon Bedrock
    
  • AWS Strands
  • CrewAI
• ‍Coding Agents: ‍
  
  • Claude Code
  • Codex CLI
  • Gemini CLI
  • OpenCode

Evaluation Without the Trust Tax

Flagging is not enforcing, and enforcing on every trace takes evaluation you can afford to run at scale. Centor, Fiddler’s evaluation engine, takes any input and automatically uses the right model to evaluate it, quickly, accurately, and at up to 98% lower cost than external LLM-as-a-judge approaches. All securely inside your environment.