Glossary

Documents & content

• Document — any saved piece of content: web page, article, note, highlight, or synced file. Has a title, excerpt, body, tags, and source.
• Library — the dashboard view of all documents in a workspace.
• Reader — the distraction-free reading view for a saved document (with editing, tagging, linking, and suggestions).
• Highlight — a selected-text save captured from the browser extension’s context menu.
• Note — a short annotation attached to a save from the extension popup.

Workspaces

• Autonomous mode — default workspace review_mode where high-authority memories and reflections with confidence >= 0.75 auto-activate, while lower-confidence reflections route to the Review Inbox.
• Context pack — the pre-flight orient returned by get_workspace_context: pinned memories, routines, topic facts, and stats.
• Manual Review mode — opt-in workspace review_mode where all agent writes land in the Review Inbox for manual approval.
• Workspace — a named, isolated container of documents and memories. Everything is scoped to a workspace.

Memory stack

• Archive — soft hide from retrieval (recoverable).
• Confidence — a memory’s reliability score (0–1); contradictions reduce it.
• Consolidation — background cron process folding repeated experience into reusable knowledge: promotes repeated episodes into procedures (episode -> procedure when >= 3 episodes share a subject) and clusters saved files into semantic facts (reference -> semantic).
• Decay — gradual weakening of unpinned memories over time.
• Draft — a memory written by an agent that has not yet been approved or auto-activated (excluded from search and retrieval until activated).
• Extraction — the automatic pass that derives tags, entities, claims, and domain from content.
• Facts (semantic) — knowledge, facts, and relationships.
• Files (reference) — imported content (how reference documents are labeled).
• Forget — permanent removal from storage and search.
• Kiomon’s decisions — the audit feed in the dashboard Review Inbox logging every autonomous write-gate action (auto-activation, merge, discard, quarantine).
• Memory — an agent-authored (or human-authored) knowledge item with a typed kind and lifecycle.
• Moments (episodic) — dated events with a timestamp and a time-to-live.
• Pin — makes a memory exempt from decay and top-ranked.
• Quarantine — safety state for low-authority claims that contradict pinned or higher-authority facts: stored and flagged with a contradicts edge, but hidden from retrieval so they never silently overwrite user facts.
• Restore — un-archive a previously hidden memory.
• Revert — lifecycle action (manage_memory action: "revert") that undoes a write-gate decision, restoring a superseded/quarantined memory or demoting an auto-activated memory back to pending.
• Review Inbox — the dashboard queue holding pending drafts awaiting review, alongside the “Kiomon’s decisions” audit feed.
• Routines (procedural) — how-to rules, triggers, and preferences agents should follow.
• Strength — a memory’s health signal, fed by retrieval and ratings.
• Write gate — the control plane pre-pass that classifies incoming drafts against active semantic memories (DISCARD, MERGE, CREATE) and routes them to auto-activation or the Review Inbox.

Retrieval

• Hybrid search — keyword matching fused with semantic (meaning-based) matching, combined into one ranked result list.
• Keyword search — exact-term matching; great for phrases, code, and technical terms.
• RAG — retrieval-augmented generation: grounded Q&A over saved documents with numbered citations.
• retrieve_memory — intent-aware retrieval tool that returns a decision-ready ContextPacket (top facts, one best procedure, episodes, evidence, warnings) ranked by task applicability (execute, plan, answer, verify, debug).
• Semantic search — meaning-based matching that finds concepts even when the wording differs.
• Snippet — a short matching passage surfaced in search results.
• Staged retrieval — answer follow-up questions from already-cited documents before doing a full-library scan.

Knowledge graph

• Knowledge graph — documents/memories as nodes, links as edges.
• Auto-linker (the Librarian) — background process that proposes related-document links with similarity scores. Pro feature.
• Link suggestion — a pending candidate link; approve it to make it real.
• Typed relation — a semantic link between memories: derived_from, supports, contradicts, related_to, precedes, follows, part_of.

Agents & protocol

• API key — the bearer credential agents use to authenticate with your account.
• get_latest_briefing — MCP tool returning the workspace’s weekly proactive knowledge digest, open commitments, and evolutions.
• MCP — Model Context Protocol; the standard for agent↔tool communication.
• MCP server — Kiomon’s hosted Model Context Protocol endpoint at https://mcp.kiomon.com/mcp, exposing the 12 memory-native tools to any MCP client.
• OpenAI-compatible API — OpenAI-shaped endpoints (chat/completions, embeddings, models) for any OpenAI SDK, n8n, or LangChain.
• record_outcome — MCP tool for task credit attribution; reports whether retrieved memories were helpful, unused, misleading, or outdated to close the learning loop.

Account & billing

• Plan — Free or Pro; determines your caps and which features are enabled.
• Trial — the 14-day Pro trial granted at signup; no card required.
• Quota — the usage caps enforced per plan (documents, storage, workspaces, keys, connectors, rate limits).

Sources & connectors

• Connector — a connected content source (Notion, Google Drive, Confluence) that syncs into your library.
• Sync — the background process that imports new and changed content from a connector.
• OAuth connection — the authorized link between your account and a provider, used by connectors.
• Source — where a document came from (extension save, highlight, note, or a specific connector).