ssreeni1

ssreeni1 / tracebase

Public

Secure local trace capture and inspection for Codex and Claude agent sessions.

28
5
100% credibility
Found May 28, 2026 at 28 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
JavaScript
AI Summary

Tracebase is a local-first observability tool for AI coding assistants that captures, encrypts, and analyzes transcript sessions from Codex and Claude Code. It stores everything privately on your computer, builds searchable indexes, serves a localhost dashboard for inspection, scores runs for quality and efficiency, and lets you compare sessions or export redacted bundles for sharing. It intentionally avoids capturing hidden model reasoning and requires explicit opt-in for any network features.

How It Works

1
📁 Your AI assistant leaves behind a trail

As you work with Codex or Claude Code, every command, file change, and decision is automatically saved as a transcript on your computer.

2
🔒 Everything stays private on your machine

Tracebase encrypts your data with a key only you have, stores it locally, and never sends anything to the internet without your permission.

3
🔍 You open a dashboard to see what happened

You start a local viewer and browse through your sessions, searching for specific events, commands, or files you want to understand better.

4
🧠 Your AI gets a report card

Tracebase analyzes each session and scores it for quality, efficiency, and risk, flagging failures, repeated commands, and wasted context.

5
You want to dig deeper or share findings
📊
Compare two runs

See exactly what improved or got worse between sessions, with token counts and failure breakdowns.

📦
Export a safe bundle

Share a redacted package that shows what happened without exposing secrets or raw data.

6
🔌 You connect it to your tools

Optional integrations let you pipe custom events into Tracebase or use it with other developer tools through a simple interface.

✨ Your AI work becomes auditable

Every session is now searchable, shareable, and understandable — you can answer questions about what your AI assistant actually did and why.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 28 to 28 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is tracebase?

Tracebase captures and inspects AI coding agent sessions locally on your machine. It imports transcript files from Codex and Claude Code, encrypts raw events with AES-256-GCM, and indexes them in a searchable SQLite database. A localhost dashboard and CLI let you explore what happened in past sessions without sending data anywhere.

Built in JavaScript with Node.js, it handles JSONL transcript parsing, provides MCP server access for AI tools, and offers Claude hook integration for live capture. The CLI exposes commands like `tracebase import`, `tracebase analyze`, and `tracebase serve` alongside utilities for export, summarization, and run comparison.

Why is it gaining traction?

Most AI observability tools require cloud services or SDK integration. Tracebase takes a local-first approach: your data stays in `~/.traces`, encrypted, never leaving your workstation unless you explicitly export it. The dashboard binds to `127.0.0.1` by default with a strict Content-Security-Policy.

The analysis engine detects failures, resteers, loops, and context waste patterns. You can compare two analyzed sessions side-by-side to see token deltas, quality scores, and risk assessments. Redacted exports are the default; raw exports require an explicit intent header.

Who should use this?

Developers running Codex or Claude Code who want to understand their agent sessions after the fact. Engineers debugging stuck or repetitive runs. Teams needing auditable traces for codebases with compliance requirements. Anyone who wants secure local storage for AI interaction logs without vendor lock-in.

Verdict

Tracebase solves a real problem with a thoughtful local-first design, but with only 28 stars and a 1.0% credibility score, it is early-stage software. The test coverage and documentation appear thorough, but it lacks the ecosystem maturity of established observability platforms. Try it on personal projects first if you need secure local jobs and trace analysis, but validate it against your security requirements before committing to production use.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.