FuHehe12

Cc Wire Analyzer — Development skill for Claude Code

Development community

Local Claude Code HTTP traffic inspector with a desktop GUI and agent API.

How to install Cc Wire Analyzer

This entry records only its repository, not the path inside it, so there is no exact command to give. Open FuHehe12/cc-wire-analyzer and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Cc Wire Analyzer does

Local Claude Code HTTP traffic inspector with a desktop GUI and agent API. Capture prompts, tools, SSE, errors, and tokens.

Alternatives in Development

Screenshot — Capture desktop, app windows, or pixel regions across OS platforms 14.6k ★
Claude Desktop Buddy — Reference and an example for the Bluetooth API for makers in Claude Cowork & Claude Code Desktop 2.5k ★
10 Context And Prompts — Prompt 10: Wire Up System Prompt, Context Gathering & Memory System 2.3k ★

README

CC Wire Analyzer — Understand Claude Code's Prompts, Thinking, and Subagent Behavior

Claude Code tells you what it did, but often you need to know why: which hidden call failed, what automatic mode is waiting for, what prompt a subagent received, or what system prompt the model actually saw.

CC Wire Analyzer records requests and responses between Claude Code and its upstream locally, then organizes that behind-the-scenes activity into a capture list, request details, a timeline, and comparable snapshots. Inspect it yourself, or let an agent search and analyze it directly.

[中文](README.zh.md) · [日本語](README.ja.md)

[Website](https://fuhehe12.github.io/cc-wire-analyzer/) · [Download the latest release](https://github.com/FuHehe12/cc-wire-analyzer/releases/latest) · [Documentation](docs/README.md) · [Changelog](CHANGELOG.md)

**Windows and macOS · Chinese, English, and Japanese UI · Optional translation and AI explanation · Recordings stored locally**

What can it help you diagnose?

Safety classifier or hidden-call errors: Find the failed request and inspect the actual upstream error, along with the related prompt, model settings, and tool calls.
Automatic mode suddenly slows down: Compare total duration and time to first token to determine whether Claude Code is thinking, retrying, or stuck on an auxiliary request such as title generation, a safety review, or token counting.
You want to see what the model actually received: Expand the complete system prompt, Messages, tool definitions, and thinking. Translate English or mixed-language content in place while keeping the original beside it.
Subagent behavior is hard to follow: Use the timeline to see who spawned whom, which prompt Task/Agent used, and what context each subagent received.
The problem is better handed to an agent: Recordings are machine-readable JSONL, and a local HTTP API is available. An agent can search, compare, and group errors without making you inspect every re