Isolated knowledge bases per venture
One private brain per project and sector — no bleed.
docs ─┐ books ─┐ podcasts ─┐ audiobooks ─┐
▼ ▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐ (one project each)
│ venture A│ │ venture B│ │ venture C│ ── structurally isolated
└────┬─────┘ └────┬─────┘ └────┬─────┘
└────────────┼────────────┘
▼
┌─────────────┐ scope → one venture (private)
│ Claude │ omit scope → span the portfolio
└─────────────┘The challenge
Running several ventures at once means several huge, distinct bodies of knowledge — documents, books, videos, podcasts, audiobooks — one per project and sector. In a single AI chat they blur together, and answers leak the wrong company's context.
How Memoception solves it
Each project and sector is a structurally isolated memory partition — a query scoped to one can only ever see that one. Isolation is enforced by the namespace + project key, not a convention.
Documents and book notes ingest as files; video, podcast, and audiobook transcripts index as history; each lands tagged to its venture.
In a chat about venture B, scope recall to that project and Claude answers only from B's knowledge — its market, its docs, its decisions.
When you want the cross-portfolio view, omit the scope and recall spans every venture at once — isolation is opt-in per query, not a wall.
memory/ingestmemory/history (transcripts)project isolationmemory/recall (scoped & spanning)MCP connector