Trace coding agent API traffic locally with Claude Tap
Run Claude Code, Codex CLI, Gemini CLI, Cursor CLI, OpenCode, and related coding agents through a local proxy so an operator can inspect prompts, tool calls, token usage, and request diffs.
npx skills add agentskillexchange/skills --skill trace-coding-agent-api-traffic-locally-with-claude-tap
Use Claude Tap when an operator needs evidence from a real coding-agent run: captured system prompts, conversation history, tool schemas, tool calls, streaming responses, token usage, and diffs between adjacent requests. It runs as a local proxy and trace viewer for supported clients including Claude Code, Codex CLI, Codex App, Gemini CLI, Cursor CLI, OpenCode, Kimi, Pi, Hermes, Qoder, Antigravity, CodeBuddy, and OpenClaw.
What this skill actually does
Invoke this instead of using the coding agent normally when the task is to debug agent behavior, compare prompt/context changes, audit tool-call payloads, or export a portable local trace for review. The scope boundary is tight: Claude Tap captures and reviews local agent API traffic for observability and debugging. It is not a general LLM gateway, hosted monitoring platform, coding agent, or SDK listing.
Inputs and prerequisites: Python 3.11+, uv or pip, and at least one supported coding-agent CLI such as Claude Code, Codex CLI, Gemini CLI, Cursor CLI, OpenCode, or OpenClaw.
Setup notes: uv tool install claude-tap or pip install claude-tap
Source and verification boundary: use https://github.com/liaohch3/claude-tap as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.