Review your daily Claude Code sessions and generate beautiful HTML learning reports. Turns AI-assisted coding into lasting knowledge.

npm install -g openrecap

Requires Node.js 18+ and access to AWS Bedrock or any OpenAI-compatible API.

# First run — interactive setup (provider, endpoint/region, model, auth)
openrecap config

# List today's sessions
openrecap

# List a specific date
openrecap --date 2026-03-14

# Generate a report
openrecap generate --date 2026-03-14

# Generate for specific sessions only (by number from list)
openrecap generate --date 2026-03-14 --sessions 1,3,5

openrecap [options]              List sessions (default)
openrecap generate [options]     Generate a learning report
openrecap config                 Re-run setup

List options:
  --date <date>       Target date or range (default: today)
                      Examples: 2026-03-14, 2026-03-10:2026-03-14
  --detail            Show estimated tokens and cache status

Generate options:
  --date <date>       Target date or range (default: today)
  --sessions <ids>    Session numbers or IDs, comma-separated (e.g. 1,3,5)
  --format <format>   Output format: html | md (default: html)
  --output <dir>      Output directory (default: ~/openrecap-reports)
  --no-cache          Skip cached analysis, re-analyze all sessions
  --concurrency <n>   LLM concurrency limit (default: 3)

1. Discover — Scans ~/.claude/projects/ for sessions matching the target date
2. Parse — Reconstructs the final conversation path from the session DAG (handles branching, compaction, sub-agents)
3. Compress — Filters noise (progress events, snapshots), truncates tool I/O, chunks large sessions
4. Sanitize — Redacts API keys, AWS credentials, JWTs, connection strings, private keys
5. Map — Analyzes each session with LLM, extracts knowledge points, tips, and problem-solution pairs
6. Reduce — Merges all session analyses into one consolidated report (auto-batches when input exceeds context limit)
7. Render — Generates a self-contained HTML or Markdown report

• Overview — Session count, projects involved, one-line summary per session
• Knowledge Points — Interactive notebook with category tabs, filterable by topic
• Practical Tips — Reusable code snippets and techniques
• Problems & Solutions — Collapsible cards: problem → cause → solution
• Further Learning — Suggested topics based on the day's work

OpenRecap supports two providers:

OpenAI-compatible APIs (recommended)

• Base URL (default: https://api.openai.com/v1)
• API Key — optional for self-hosted endpoints
• Model ID — e.g. gpt-5.4,

AWS Bedrock

• AWS Region (default: us-east-1)
• AWS Bearer Token (AWS_BEARER_TOKEN_BEDROCK) — or standard AWS credential chain

Session data is sent to your configured LLM provider for analysis.

Safeguards:

• Explicit consent prompt during setup
• Best-effort secret redaction before any LLM call
• Reports and cache stored locally only (~/.openrecap/)

git clone https://github.com/founddream/openrecap.git
cd openrecap
bun install

# Run directly (no build needed)
bun run src/index.ts

# Tests
bun test

# Type check
bun run typecheck

# Build & run
bun run build
./dist/index.js

MIT