ClaudeMap

TraceCrate

A privacy-first, client-side trace workbench — turn Claude Code, Codex and OTLP logs into a searchable timeline, recorded metrics and side-by-side comparisons. No backend, no telemetry.

SDKs & Toolingobservabilitytracingdebugginglocal-firstchinese
Visit projectFankChen/tracecrate

TraceCrate reads trace files you choose — Claude Code sessions, Codex logs, OTLP exports — and turns them into a searchable timeline (event-kind filters, relative times, tool inputs/outputs, metadata, 100 events per page), recorded metrics, heuristic diagnostics and side-by-side run comparisons. Its philosophy is evidence, not invented certainty: tokens, explicit errors and elapsed time are reported when present, missing values stay unknown, and there are no cost estimates. It reads traces but never runs agents or recorded commands, never scans agent directories, and exports to JSON or a standalone HTML report with a structure-only redacted preview you review before sharing. A hosted demo runs on synthetic data. v0.1.0, bilingual, live on GitHub Pages.

What it does

  • Drop Claude Code, Codex or OTLP trace files and get a searchable timeline with event-kind filters, tool inputs/outputs and metadata, 100 events per page
  • Compare runs side by side and read heuristic diagnostics — explicit tool errors, long tool durations — built on recorded evidence, not invented certainty
  • Export JSON or a standalone HTML report, with a structure-only redacted preview to review before anything leaves your machine
  • No backend, no telemetry, no accounts, no API keys — it reads traces, never executes recorded commands, and never scans agent directories; you pick every file

When to use it

Answering "what did the agent actually do" after a long session — sequence, tool failures and timing diffs between two runs, without uploading anything to anyone.

Getting started

# local (Node.js >= 22.12, 24 recommended):
git clone https://github.com/FankChen/tracecrate.git
cd tracecrate
npm ci
npm run dev

# or open the hosted demo (synthetic data, nothing to install):
# https://fankchen.github.io/tracecrate/

Things to know

v0.1.0, a fresh first release. Strictly a viewer: it doesn't run agents, and it deliberately reports unknown values as unknown — no cost estimates, so don't expect spend dashboards.

Repository

FankChen/tracecrate

Charted

Related on the map

  • Harness Bridge

    SDK / Tool

    A local launcher that points Claude Code and seven other coding harnesses at any OpenAI-, Anthropic- or Responses-compatible endpoint — without editing a single config file.

    • launcher
    • local-models
    • cli
    • menu-bar
    View

  • Claude for Siri

    SDK / Tool

    Answers macOS Spotlight and Siri prompts with your signed-in Claude Code account — an experimental macOS 27 model-delegation provider.

    • macos
    • siri
    • spotlight
    • model-delegation
    View

  • AI Coding Assistant Data Extractor

    SDK / Tool

    Extract your own local chat history from ten AI coding assistants into one normalized JSONL — for fine-tuning, analytics, or backing up years of conversations. Standard library only.

    • cli
    • data-export
    • jsonl
    • backup
    View