Most developers spend hours building products, but when something breaks, they end up spending even more time inspecting raw network traffic and digging through endless JSON payloads.
That's why we built Custom Viewers in NetworkSpy.
Instead of forcing your team to interpret the same traffic differently every time, Custom Viewers let you transform raw requests and responses into visualizations tailored to your application and workflow.

Mailwarm
Can NetworkSpy capture and visualize the full agent loop, like prompts, responses, and function calls, in one timeline?
NetworkSpy
Hi @thamibenjelloun, thanks for the question!
NetworkSpy's built-in LLM viewers are designed for OpenAI-compatible /completions and /chat/completions endpoints. They provide a structured, developer-friendly view of requests and responses, including messages, available tools, response choices, and streaming chunks. The goal is to make debugging intuitive and enjoyable, rather than forcing developers to sift through large volumes of raw JSON.
For full agent-loop visualization—such as prompts, responses, tool/function calls, and execution flow in a single timeline—the built-in viewers are not intended to serve as a backend observability solution.
That said, NetworkSpy's Custom Viewer Builder gives you the flexibility to create your own visualization layer. You can build viewers for individual traffic inspections or aggregate multiple traffic inspections into a single custom view. This allows you to parse traffic data and present it in whatever format best fits your workflow, including agent execution timelines and cross-request flows.
Custom viewers can also be versioned in GitHub and shared across your team, ensuring everyone has a consistent understanding of how data and requests move between clients, servers, and AI agents.