A community-driven registry for the Claude Code ecosystem. Not affiliated with Anthropic.
Are you the author? Sign in to claim
Precomputed reasoning cache for AI infrastructure decisions — 220K+ pages tracking the open-source AI ecosystem
PT-Edge is a precomputed reasoning cache for AI infrastructure decisions. It tracks 220,000+ AI repos across GitHub, PyPI, npm, Docker Hub, HuggingFace, and Hacker News, scores them daily on quality, and publishes the results as a 220,000+ page directory site.
The site serves two audiences: AI agents reading pages on behalf of humans (structured, front-loaded, machine-readable) and humans reading directly (navigable, trustworthy, original analysis). Every page is designed so an AI agent can land on it and walk away with a confident, citable recommendation in one pass.
Every major AI lab's crawl infrastructure treats the site as a primary data source. The access logs are themselves an intelligence layer — see Demand Radar below.
Directory site: mcp.phasetransitions.ai — 220,000+ pages across 17 domains with 2,400+ categories, updated daily.
Built by Graham Rowe
The entire system runs on a single server instance for under $300/month.
| Dimension | Max | Signals |
|---|---|---|
| Maintenance | 25 | Commit activity (30d), push recency |
| Adoption | 25 | Stars (log scale), monthly downloads, reverse dependents |
| Maturity | 25 | License, PyPI/npm packaging, repo age |
| Community | 25 | Forks (log scale), fork-to-star ratio |
Tiers: Verified (70-100), Established (50-69), Emerging (30-49), Experimental (10-29)
Every bot hit on the site is latent intelligence. The access logs carry three layers of signal:
The Demand Radar extracts these signals and feeds them into content prioritisation — eventually via trained ML models rather than hand-tuned weights. See scratch/demand-radar/ for the full analysis.
Python, FastAPI, PostgreSQL + pgvector, LLM enrichment (multiple providers), static site generation via Jinja2 + Tailwind CSS. Hosted on Render. MCP tools and REST API for programmatic access.
This is a production system with no staging environment. The database is a live 1GB+ PostgreSQL instance — queries hit real data. See docs/development.md for setup notes and safety rules.
docs/strategy.md — strategic positioning and reasoningdocs/roadmap.md — phased build plandocs/design/worker-architecture.md — task queue and worker designdocs/development.md — development setup and database safetyscratch/demand-radar/ — access log intelligence analysisMIT — see LICENSE.
Run Claude Code as an MCP server so any agent can delegate coding tasks to it
Browser automation using accessibility snapshots instead of screenshots
MCP server integration for DaVinci Resolve Studio
@designcomputerSecure MCP server for MySQL database interaction, queries, and schema management
