Glossary
Purpose: Align vocabulary for internal teams, prospects, and AI prompts.
| Term | Definition |
|---|---|
| MCP (Model Context Protocol) | Standard interface many AI clients use to connect tools; Reflect exposes memory operations over MCP. |
| Ambient Mode | User preference (default On). On = silent retrieval + silent capture of durable facts. Off = Suggested Memories (approval-required). |
| Suggested Memories | Ambient Off path: candidates are staged for dashboard approve / edit / reject. |
| Visibility / allowed vendors | Per-memory scoping so each connected AI tool only sees what it is permitted to see. |
| Shared memory | Organization pool or sub-team pool after an explicit share. The two pools do not overlap. Ambient captures stay personal until shared. |
| Organization pool | Company-wide shared memories. MCP: read_org_memories. Default scope on share_memory. |
| Sub-team pool | Narrower pool for one team inside the org. MCP: read_team_memories with share_memory scope team. |
| Project folder | Named folder under personal, org, or sub-team scope. Stable proj_ tag. List with list_projects. |
| Supersession / version history | Newer truth replaces or updates prior content while history remains available for audit and context. |
| Session briefing | On-connect context that orients the model to the user’s memory state and Ambient preference. |
| Deployment modes | Hosted, isolated-hosted, and self-host - differing ownership and network boundaries. |
| Audit trail | Record of reads, writes, auth, and admin actions (including Ambient captures) for compliance review. |
| Async diligence | Public markdown + prompts + PDFs that let prospects evaluate Reflect with their own AI before a live call. |
| NDA technical pack | Deeper architecture, questionnaires, and competitive materials shared only for active evaluations. |