Skip to main content

Documentation Index

Fetch the complete documentation index at: https://whyops.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

WhyOps is a decision infrastructure platform for autonomous agent systems. We move beyond basic prompt and token logging to help you understand exactly why your agents make the decisions they do. WhyOps runs in a proxy architecture that sits seamlessly between your application and your model providers.
WhyOps is currently deployed in production. You can proxy requests, evaluate agent runs, and view decision graphs in real-time.

​
The Problem

Traditional observability tools fail when it comes to autonomous AI agents because they track deterministic execution paths (Request → Response). Agents, however, operate using:
  • Non-deterministic planning loops
  • Internal chain-of-thought and invisible reasoning
  • Tool calling loops and failure retries
  • Dynamic context assembly
When an agent fails, you need to answer:
  • Why did it invoke that specific tool?
  • Why did it enter an infinite retry loop?
  • How did the system prompt influence this specific chain of thought?

​
The WhyOps Solution

WhyOps intercepts LLM traffic to trace these internal decision loops automatically. We do this primarily through our Proxy Layer that captures:
  1. Complete LLM Interactions: Prompts, tool calls, and LLM responses.
  2. Invisible Signatures: A mechanism that injects hidden tokens to trace context across disjointed API requests, stitching them into a unified agent “Thread.”
  3. Agent Evaluations: Our whyops-analyse service runs evaluations across traces, scoring runs on dimensions like intent precision, tool routing quality, and reliability.

​
Key Features

Official SDK Packages

Use @whyops/sdk, whyops, or the Go module to combine agent init, proxy patching, and manual runtime events in one integration surface.

Invisible Signatures

Automatically groups stateless LLM calls into coherent traces without requiring complex client-side SDK instrumentation.

Automated Static Analysis

Detects duplicate tool calls, missing responses, missing context, and infinite retry loops passively.

Proxy Architecture

Drop-in replacement for OpenAI and Anthropic API endpoints. Just change your Base URL.

Decision Graph UI

Visualize the precise causal chain of reasoning steps that led to an agent’s final answer.