Learning and knowledge
Smart Knowledge — capture what you learned once, recall it in the next session.
Official recipe: Knowledge Keeper
Courses, docs, and tutorials tempt you to re-open the same tabs every week. Knowledge Keeper fetches long-form sources, dedupes redundant sections, and stores structured facts in team-scoped memory — so the next chat starts from what you already verified, not from a blank search box.
What feels broken
- Tutorial tabs get re-fetched instead of remembered.
- Notes scatter across chats with no durable scope.
- Highlight reels from search snippets are not the same as learned facts.
How SmartGate shapes it
- Fetch the canonical doc or course page through the gateway (clean Markdown body).
- Dedup overlapping sections before summarizing.
- Write distilled facts to team memory after sign-in — not world knowledge, your scope.
- Recall memory in the next session instead of re-running the same searches.
What you get
- Personal knowledge base without maintaining a second app.
- Less redundant fetch traffic on material you already studied.
- Learning loops that respect token caps and gateway audit.
Gateway path after sign-in (for integrators): smart_search → smart_fetch → smart_memory → smart_pipe
Related roles and scenes
Same gateway, different jobs — pick the path that matches how you work.
- Research and decisionsPersonal scene: search, fetch, dedup, and compress before the model answers.
- MCP for research and decisionsPersona guide: one MCP server entry, then governed research on every question.
- FinOps leadOwn the AI bill: per-team quota, alerts, and audit rows you can sign off on.
- Agent buildersWire one Streamable HTTP MCP endpoint — seven smart_* primitives, one runtime.