Since you're using Claude, Save money on LLMs tokens and have accurate implementation.

If you are using Cursor or Claude Code, you must try this open-source tool, it will save you MONEY and TIME.

If you're building complex projects and your context keeps getting larger, losing track of what's current and what's outdated, you're going to love this.

🚨 THE PROBLEM:

When using LLMs to build products, we generate tons of context—requirements, specs, design docs. But then:

❌ Can't tell which document is the latest/authoritative

❌ No traceability between layers (business → system → code → tests)

❌ Upstream requirements change, downstream doesn't get updated

❌ Your LLM uses STALE context → incorrect output

❌ Burning tokens on entire files when you need specific snippets

✅ THE SOLUTION: contextgit

A tool built specifically for LLMs to trace context efficiently.

Instead of dumping entire files into context, contextgit lets your AI navigate requirements like a graph database—extracting only what's needed.✂️ Extract ONLY relevant snippets (save tokens = save money)

🔗 Track relationships: business → system → architecture → code → test

🔍 Auto-detect staleness via checksums

🤖 Full JSON output for LLM integration

📊 See project health: orphaned requirements, broken links, stale items

💡 KEY USE CASES:

🔹 LLM-driven development: AI asks for requirements by ID, not entire docs

🔹 Massive token savings: Feed 20 lines instead of 5,000

🔹 Auto-detect stale context: LLMs know when requirements are outdated

🔹 CI integration: Block PRs with stale requirements

 📈 REAL IMPACT:

Instead of: LLM reads 5000-line docs → wastes tokens

Now: contextgit extract SR-010 → 20 lines → saves hundreds in API costs

⭐ If you find this useful, please star the repo!

100% open source (MIT), production-ready (v1.0.1). This is the missing piece for LLM-driven development.