Item detail

liaohch3/claude-tap

liaohch3/claude-tap is an MIT-licensed local traffic inspector for coding-agent APIs that intercepts, records, and replays HTTP traffic from Claude Code, Codex CLI, Gemini CLI, Cursor CLI, OpenCode, Kimi/Kimi Code, Pi, and Hermes in a local trace viewer, so AI builders can debug, audit, and reproduce the actual API calls a coding agent makes instead of guessing from prompts and stdout.

Score8.0
Popularity7.5
Risklow
TierGold
Score breakdown
Usefulness9.0
Novelty8.0
Momentum7.0
Maturity6.5
Open-source/build8.4
Evidence7.2
Workflow potential9.5
Setup ease8.8

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Useful for AI builders, agent platform engineers, and security/QA teams who need to see exactly what a coding agent sent to the model and what came back, because Claude-Tap captures the full request/response stream in a local viewer, which means regressions, prompt-injection probes, and silent model swaps become visible and reproducible rather than guesswork from logs.

Who should use it

AI builders who want to see the actual HTTP traffic between a coding agent and the model APIAgent platform engineers debugging regressions in Claude Code, Codex, Gemini, or CursorSecurity and QA teams running prompt-injection probes against coding agents in a reproducible way

Who should skip it

Skip if the source link, docs, or setup requirements do not match your workflow.

Risk explanation

It records full API traffic from coding agents including prompts and tool calls, so keep the trace viewer local, redact any captured secrets before sharing traces, and do not run it on workloads that mix production credentials with dev traffic.

Evidence links

Closest alternatives / related signals

agent-debuggingagent-observabilityapi-debuggingclaude-codecodexgemini-clicursortrace-viewer