@seizn/author-mcp-server

Search Seizn memory graph by semantic query.

Recall canon facts for a named entity (character / location / object / rule / promise / event). Returns the canonical name, last 3 mentions with chapter snippets, current state, and any pending conflicts. Use this when the user asks about a name they wrote earlier and may have forgotten.

Check a passage of prose for canon conflicts (character contradictions, timeline issues, broken promises, rule violations). Returns conflicts with severity (P1 critical, P2 warning, P3 stylistic) and rationale. Use this on a finished scene before committing to draft.

Approve a fact as canon for an existing entity. Use when the user confirms an AI-suggested fact ("yes Seoyun is a reporter as of Ch.4") or when adding a new canonical fact mid-conversation. The fact becomes the source of truth for future recall and conflict checks.

Semantic search across all entities in a project. Use when the user asks about a concept rather than a specific name (e.g. "what happens at the temple in chapters 3-7", "characters with grief arcs"). Returns entities ranked by semantic relevance.

Return chapter-by-chapter beats for the project, optionally filtered by chapter range. Use when the user asks "what happened between Ch.5 and Ch.7" or needs to see story progression.

Return a relationship graph subset rooted at a given entity. Includes direct ties (1-hop) and indirect (2-hop) with edge kinds (rival, ally, family, etc). Use for character-relationship questions like "who is Seoyun connected to".