{"id":"cortex-works-cortex-scout","name":"cortex-scout","af_score":70.8,"security_score":46.5,"reliability_score":30.0,"what_it_does":"CortexScout is a self-hostable Rust engine for web search and web extraction for AI agents. It exposes tools over an MCP stdio server (and optionally an HTTP server) and includes stateful browser automation (CDP) plus optional Human-in-the-Loop (HITL) flows for handling bot challenges and logins.","best_when":"You need an agent-facing, self-hosted MCP toolset for search + extraction, including stateful automation and optional HITL to handle hard-to-fetch JS-heavy and auth-gated pages.","avoid_when":"You need a simple, stable public API with no browser automation or you must avoid interactions that resemble bypass/automation on protected targets; also avoid in environments where outbound scraping/automation is prohibited.","last_evaluated":"2026-03-30T15:18:38.213416+00:00","has_mcp":true,"has_api":true,"auth_methods":["MCP stdio (local process) usage; environment-variable configuration for optional LLM API key and optional HTTP server runtime"],"has_free_tier":false,"known_gotchas":["HITL/profile flows may cause Chromium profile lock errors if run concurrently or if the same profile is reused while Chrome/Brave is open.","Using too-verbose RUST_LOG (info) can flood stderr and confuse MCP clients; recommended RUST_LOG=warn.","Proxy support is optional and enabled only when IP_LIST_PATH is set; leaving it unset disables proxy tools.","HITL/browser steps that require visible interaction need a desktop session and an --all-features build."],"error_quality":0.0}