yolog.dev Desktop

Product Introduction

1. Definition: yolog.dev Desktop is a local-first desktop application (macOS) that archives, indexes, and analyzes AI coding sessions, specifically targeting conversations generated by tools like Anthropic's Claude Code. It transforms ephemeral JSONL session files into a structured, searchable SQLite database on the user's machine.
2. Core Value Proposition: It solves permanent loss of AI coding session context by providing offline replay, search, and analytics for developer interactions with AI coding assistants, enabling retrospective analysis of technical decisions, cost tracking, and productivity insights without cloud dependencies.

Main Features

1. Session Replay & Playback:
   How it works: Imports Claude Code JSONL session files and provides a frame-by-frame timeline playback with keyboard controls (Space/Arrows). Supports Cinematic Mode for full-screen, ambient-effect-enhanced viewing. Enables precise review of AI-generated code suggestions and debugging steps within the original conversational context.
2. Search & Indexing:
   How it works: Performs full-text search across all archived sessions using local indexing (SQLite). Highlights matches within messages, enabling developers to instantly locate specific decisions, error messages, or code snippets (e.g., "OAuth redirect fix") across potentially thousands of messages/hours of sessions.
3. Vibe Metrics & Analytics:
   How it works: Computes quantitative metrics like Token Leverage (user tokens vs. AI tokens), Burn Rate (cost accumulation speed), Flow Detection (based on interaction patterns), and Streak Tracking (productive session sequences). Provides a Vibe Score aggregating session quality indicators.
4. Cost Tracking:
   How it works: Calculates estimated session and project costs based on token usage (input/output) derived from session data. Offers per-project/per-session cost breakdowns, crucial for budgeting AI tool usage (e.g., "$131.32 total cost, 140.5M tokens used").
5. Auto-Sync & Local-First Architecture:
   How it works: Monitors a user-specified directory (e.g., ~/.claude/projects/) for new Claude Code JSONL files. Automatically imports and indexes new sessions in real-time. All data (session content, indexes, metrics) resides locally in SQLite (~10MB footprint). Operates fully offline; no data leaves the user's device.

Problems Solved

1. Pain Point: Irretrievable AI coding session history. Developers lose critical context, decisions, and debugging steps buried in transient JSONL files that are rarely revisited manually, leading to repeated work and lost insights.
2. Target Audience: Full-stack developers, AI engineers, and engineering teams using Claude Code (or similar AI coding tools) for daily tasks like debugging, feature implementation, or code reviews. Particularly valuable for solo developers and small teams needing cost visibility.
3. Use Cases:
   • Replaying the exact steps taken with an AI assistant to fix a complex bug (e.g., OAuth redirect issue) weeks later.
   • Searching all past sessions to find where a specific algorithm or API integration was discussed.
   • Auditing AI tool costs per project to optimize budget allocation.
   • Onboarding new team members by showing recorded AI-assisted problem-solving sessions.

Unique Advantages

1. Differentiation: Unlike manual note-taking or cloud-based session trackers, yolog.dev offers privacy-first, offline-native session archaeology specifically designed for the structure of AI coding tools (JSONL). It surpasses basic chat history by adding deep metrics, search, and replay absent in native Claude interfaces.
2. Key Innovation: Its automatic JSONL-to-SQLite indexing engine combined with local-only analytics computation (Vibe Score, Token Leverage, Cost Tracking) provides powerful insights without compromising data privacy or requiring internet connectivity. The Cinematic Mode playback offers a uniquely focused review experience.

Frequently Asked Questions (FAQ)

1. Does yolog.dev Desktop work with AI tools besides Claude Code?
   Currently, yolog.dev Desktop has native integration for Claude Code session JSONL files. The developers state support for additional AI coding tools is planned for future updates ("more tools support coming soon").
2. Is my session data sent to the cloud when using yolog.dev?
   No. yolog.dev Desktop operates on a strict local-first, privacy-first principle. All session data, indexes, and analytics are stored exclusively in an SQLite database on your macOS machine (~10MB). The app functions entirely offline; no session content is transmitted externally.
3. How does yolog.dev calculate costs for my AI sessions?
   yolog.dev estimates costs by analyzing token counts (input and output) within your Claude Code JSONL session files and applying known or estimated pricing models for the AI model used. It provides per-session and per-project cost breakdowns based on this local analysis.
4. Can I use yolog.dev to search across multiple projects?
   Yes. Once you point yolog.dev to your root directory containing Claude Code projects (e.g., ~/.claude/projects/), it automatically indexes sessions from all sub-projects. The global full-text search functionality allows you to find terms across every indexed session in every project.
5. What macOS versions are supported, and is there a Windows/Linux version?
   The current download is specified for macOS. The website does not list specific version requirements or announce immediate availability for Windows or Linux. Users should check the official yolog.dev download page for the latest platform support information.