{"id":"juyterman1000-entroly","name":"entroly","af_score":51.5,"security_score":30.2,"reliability_score":32.5,"what_it_does":"Entroly is a context engineering/optimization tool for AI coding agents. It ingests and indexes a workspace, scores/selects and compresses codebase context to fit within token budgets, and delivers that optimized context to agents via an MCP server or a local HTTP proxy. It also provides a dashboard/metrics, explain endpoints, and optional reinforcement-learning-style weight adjustment from feedback.","best_when":"You want local, IDE-agent-friendly context optimization across an entire repository, especially when agents only see a small subset of files.","avoid_when":"You cannot expose localhost endpoints to your agent tool, or you need a documented, standards-based API contract (OpenAPI) with strong security controls beyond local dev defaults.","last_evaluated":"2026-03-30T13:50:04.742420+00:00","has_mcp":true,"has_api":true,"auth_methods":["Localhost integration via MCP (tool registration)","Local HTTP proxy on localhost (no auth described in README)"],"has_free_tier":false,"known_gotchas":["Proxy endpoints are described as localhost services; agents must be configured to point to the correct base URL (e.g., http://localhost:9377/v1)","Port conflicts can occur (README mentions 9377 already in use)","MCP/IDE integration relies on `entroly init` generating appropriate config/artifacts (e.g., .cursor/mcp.json for Cursor)","Rust engine may not load; falls back to Python if native wheel/tooling unavailable"],"error_quality":0.0}