- Brain setup + conversational onboarding
- Three-layer context (L1/L2/L3)
- Signal-driven auto-persistence (auto/manual)
- Conversation distillation
- AI-initiated project memory setup
- Project-owned local persistence (.loci/memory.md + .loci/profile.md + .loci/progress/ + .loci/decisions/)
- Brain project index (projects/index.md) instead of memory warehousing
- Cross-terminal awareness (hooks)
- Dashboard (local Node server,
node .loci/dashboard/server.js) - Multi-tool compatibility (adapt.sh)
- Growth tracking (evolution.md)
- Extension modules on demand
- Memory Consolidation（记忆整合） — 模拟海马体睡眠整合机制：
- 每天首次对话时，LLM 回顾过去 24h 的蒸馏结果
- 发现跨领域模式（"你最近三个决策都在简化架构"）
- 生成洞察追加到
me/insights.md或me/evolution.md - 灵感来源：Google Always-On Memory Agent 的 Consolidation 阶段
- 可手动触发（
/loci-consolidate）或自动（每日首次对话）
- Source Citations（来源引用） — 蒸馏时标注来源，查询时引用
- CI test suite (GitHub Actions, macOS + Ubuntu matrix)
- Search across brain (full-text grep + fuzzy match)
- Project memory maintenance for large
progress/and completed project todos - Multiple routing modes (open, manual, silent)
- Privacy boundary configuration (blocked_tags, custom rules)
- Distillation level presets (verbose/balanced/minimal)
- Retention policy (auto-archive after N days)
- Nested @import support in adapt.sh
- Cloud sync (git remote based, free tier = own repo, paid = managed)
- Embedding-based semantic search
- Mobile-friendly dashboard (PWA)
- Team/multi-user brain with permissions
- Plugin system for custom extensions
- Native support for Cursor/Windsurf/Cline (beyond adapt.sh)
- AI memory API (commercial offering)
- Shareable brain stats card (/loci-stats)
- Multi-modal inbox (文件监听目录，自动摄取非 markdown 内容)
- More example brains (designer, student, researcher)
- Brain health check script
- Improved install.sh (conflict detection, dry-run mode)
- Troubleshooting guide + FAQ
- Landing page
- Google Always-On Memory Agent (2026-03) — 通用 Agent 记忆后端，SQLite + Gemini，三阶段流水线（Ingest → Consolidate → Query）。Loci 差异：身份层 + 多项目编排 + 零基础设施。
- Mem0 — API-first 记忆服务，向量数据库 + embedding。Loci 差异：纯 markdown、本地优先、透明可审计。
- Letta (MemGPT) — 自管理记忆的 Agent 框架。Loci 差异：不需要运行时，寄生在宿主工具中。
- ChatGPT Memory — 平台内置，扁平列表。Loci 差异：结构化三层、可导出、不锁定。
These were discussed and designed but deliberately cut from v1.0 to ship faster:
- Routing modes were fully designed (see git history) but only tag-routed ships in v1.0
- Compression was designed (>200 lines, >30 days → summarize) but removed for simplicity
- Privacy boundaries were designed (blocklist mode) but not exposed in v1.0
- Distillation levels were designed (verbose/balanced/minimal/custom) but v1.0 uses balanced only