
I kept wasting time googling
"what changed in Claude 3.7" or
"how does Cursor compare to Windsurf"
and getting scattered blog posts.
So I built agentcodex.dev
Each agent gets a Wikipedia-style page with:
→ Full version history (what changed and when)
→ Capability scores across 6 dimensions
→ Side by side comparisons
→ Daily auto-updates via a pipeline that reads
official changelogs and extracts new versions
Currently tracking 20 agents including Claude,
ChatGPT, Copilot, Cursor, Gemini, Windsurf,
Mistral, Grok and more.
Tech stack: Next.js 15, Supabase, Vercel,
Claude API for extraction, GitHub Actions for
daily automation
Looking for feedback on:
Happy to answer questions about the tech stack or the auto-pipeline approach.
The most interesting part was building a deduplication system using content hashing so the daily pipeline doesn't reprocess the same articles.
Also open to suggestions on which agents to add next!