What we shipped at
@cognee_ last week
→ Cloud pricing dropped to $1 per 1M tokens
→ LLM-free graph extraction landed (GLiNER) - build knowledge graphs without burning tokens
→ Coding-agent plugins (Claude Code, Codex) got faster recall: concurrent scope dispatch + per-agent identities with scoped permissions
→ Activity log now shows per-operation cost, and memory coverage scores only what was actually measured
→ Docs overhaul: fixed API playground, Rust & TypeScript SDK tabs, new pipeline demos
151 issues closed across the team. Shipping memory that works.