Rudel is an open-source analytics platform designed to provide visibility into Claude Code sessions, tracking metrics like efficiency, abandonment rates, skill usage, and task success. The makers found surprising insights from their own data, highlighting a general lack of understanding in AI agent performance. It aims to help developers and teams optimize their AI coding tool usage.
A starter prompt for Claude Code, what you'll need, and how to reach them.
You are an expert full-stack developer. Create a Next.js 16 App Router application with React 19, Tailwind v4, and a Neon Postgres database for storing and analyzing Claude Code session data. The core functionality is to provide analytics on AI coding sessions, including session efficiency, abandonment rates, skill usage, and task success.
**Phase 1: Data Ingestion (MVP)**
1. **Frontend**: Create a simple Next.js page that can receive JSON data representing a Claude Code session (e.g., via a POST request from a browser extension or manual upload). Define a clear schema for session data: `sessionId`, `userId`, `startTime`, `endTime`, `taskType` (e.g., 'refactoring', 'documentation'), `interactions` (array of objects: `prompt`, `response`, `tokenCount`, `timestamp`, `isSkillUsed`).
2. **Backend (API Route)**: Implement a `/api/sessions` Next.js API route that accepts POST requests with session data. Validate the incoming JSON against the defined schema.
3. **Database**: Design a Postgres schema to store this session data. Use Prisma for ORM. Model `Session` and `Interaction` tables. Ensure `userId` is indexed for future user-specific analytics.
4. **Persistence**: Save the validated session data into the Neon Postgres database.
**Phase 2: Basic Analytics (MVP)**
1. **Analytics Service**: Create a backend service (e.g., a simple utility function or a dedicated API route) to process raw session data and calculate basic metrics:
* Total number of sessions.
* Average session duration.
* Total token count.
* Percentage of sessions where skills were used (based on `isSkillUsed`).
* Simple abandonment rate (e.g., sessions ending within the first 60 seconds).
2. **Dashboard (MVP)**: Build a simple React component to display these aggregated metrics. Use basic Tailwind CSS for styling. Show counts, percentages, and average values on a single page.
**Phase 3: Verify**
* Ensure that JSON session data can be successfully posted to the API route and stored in the Neon Postgres database.
* Verify that the basic analytics metrics are calculated correctly and displayed on the dashboard for the ingested data.Reach heavy users of AI coding tools and AI engineers to offer Rudel as a deeper analytics solution to complement AI-usage-monitor or as a real-world session capture for agent-eval-lab.
We built rudel.ai after realizing we had no visibility into our own Claude Code sessions. We were using it daily but had no idea which sessions were efficient, why some got abandoned, or whether we were actually improving over time. So we built an analytics layer for it. After connecting our own sessions, we ended up with a dataset of 1,573 real Claude Code sessions, 15M+ tokens, 270K+ interactions. Some things we found that surprised us: - Skills were only being used in 4% of our sessions - 26% of sessions are abandoned, most within the first 60 seconds - Session success rate varies significantly by task type (documentation scores highest, refactoring lowest) - Error cascade patterns appear in the first 2 minutes and predict abandonment with reasonable accuracy - There is no meaningful benchmark for 'good' agentic session performance, we are building one. The tool is free to use and fully open source, happy to answer questions about the data or how we built it.
Reply in the HN thread to keks0r's comment
“I'm a solo operator building developer tools and I'm very interested in the insights from Rudel. I've built a quick prototype for capturing and analyzing basic Claude Code session data and would love to collaborate or share my approach.”
Open the original ↗