Writing MCP

Run read-only diagnostics for a generated project backup bundle. Reports missing, partial, wrong-project, incompatible-schema, tampered, unreadable, and stale backups without mutating SQLite or generated files.

Run read-only structure diagnostics against the current index, sync files, and generated structure exports. Reports canonical drift, ambiguous folder-derived structure, unknown chapter links, epigraph conflicts, compatibility chapter mismatches, and export trust/staleness issues without repairing files or mutating the database.

Discard a pending proposal without applying it. The proposal is deleted and the prose remains unchanged.

Re-derive lightweight scene metadata from current prose (logline and character mentions) and clear metadata_stale for that scene. Only available when the sync dir is writable.

Start an asynchronous prose-derived relationship repair job that infers scene character mentions and updates scene character links. Defaults to dry_run=true; apply mode retains scene sidecar characters as compatibility output, then syncs SQLite scene_characters and refreshes project backups for changed scenes.

Generate a deterministic project backup bundle from SQLite canonical state. Writes manifest.json, canonical.snapshot.json, and operations.jsonl under WRITING_SYNC_DIR for explicit review and future restore workflows; this is generated transparency only and does not mutate canonical state.

Generate a deterministic JSON structure export from SQLite canonical state for Git review and future explicit recovery workflows. The export is generated transparency only: editing it does not change canonical state.

List canonical epigraphs for a project, optionally narrowed to a canonical chapter_id or read-scope numeric chapter alias.

Find scenes by filtering on character, Save the Cat beat, tags, canonical chapter identity, numeric chapter alias, or POV. Returns ordered scene metadata only — no prose. Most filters are optional and combinable. `character` and `pov` prefer character_id values and may resolve unambiguous project-scoped character names when project_id is provided. `tag` and `beat` match case-insensitively when unambiguous; near matches are suggestion-only. `chapter_id` requires `project_id` and accepts exact canonical IDs or unambiguous case variants; ambiguous case variants return candidate IDs. Numeric `chap…

Attach a continuity or review note to a scene as compatibility review metadata. Flags are not canonical relationship authority and do not mutate SQLite; use audit_relationship_metadata, connect_character_place_evidence, record_character_relationship_beat, or link_reference_evidence when the note identifies relationship repair work.

Get every scene a character appears in, ordered by part, canonical chapter order, and scene position. Returns scene metadata only — no prose. Use this as the primary structural entry point when the question is about a character's progression through the manuscript. Supports pagination via page/page_size and auto-paginates large result sets with total_count. Use list_characters only when you need help finding a character_id. Response shape note: always returns a structured envelope (`results`, `total_count`, with pagination fields when paging is active).

Get status and result for an asynchronous job started by async tools such as import_scrivener_sync_async, merge_scrivener_project_beta, or enrich_scene_characters_batch. Use this to poll job progress after receiving a job_id. Common next step: if status is still running, call this tool again; if status is completed inspect result, and if status is failed or cancelled inspect job/result diagnostics.

Description is built at runtime.

Get full character details: role, arc_summary, traits, the canonical sheet content, and any adjacent support notes when the character uses a folder-based layout. Use this when the reasoning task needs the character's canonical profile rather than only their scene progression. Response shape note: returns a structured envelope (`results`, `total_count`) with one result row.

Get full place details, including canonical sheet content plus retained sidecar associated_characters and tags as compatibility/review notes. Use connect_character_place_evidence when scene-backed character/place evidence is paired, or connect_scene_place_evidence when scene evidence is place-only. Response shape note: returns a structured envelope (`results`, `total_count`) with one result row.

Resolve prose-styleguide.config.yaml with cascading precedence (sync root, then universe root, then project root). Applies language-derived defaults and nested quotation defaults when omitted.

Get metadata for a reference document by doc_id. Optionally includes exactly one hop of related reference docs.

Show how the relationship between two characters evolves across scenes, in order. Uses explicitly recorded relationship entries — returns a NO_RESULTS error if no entries exist yet. Use list_characters to get character_id values. Response shape note: returns a structured envelope { results, total_count, from_character, to_character }.

Show the active runtime paths and capabilities for this server instance (server version, sync dir, database path, writability, permission diagnostics, and git availability). Use this to verify which manuscript location is currently connected.

Load the full prose text of a single scene. Use this for close reading, continuity checks, or when you need the actual writing. For overview or filtering, use find_scenes instead — it is much cheaper. Optionally retrieve a past version from git history. If scene IDs are reused across projects, omitting project_id returns CONFLICT with candidate project_ids.

Get ordered scene metadata for all scenes belonging to a thread, including the per-thread beat. Returns a structured JSON envelope with thread metadata, results, and total_count. Use this when the question is about subplot movement, continuity, or recurring storyline structure across scenes. Supports pagination via page/page_size.

[STABLE] Import Scrivener External Folder Sync Draft files into this server's WRITING_SYNC_DIR by generating scene sidecars and reconciling by Scrivener binder ID. This is the recommended default path for first-time setup before sync().

[STABLE] Start an asynchronous Scrivener External Folder Sync import job. This is the recommended default import path when the sync tree is large. Returns immediately with a job_id to poll via get_async_job_status.

Link scene, character, place, or reference evidence to a reference document. This is the outcome-level workflow for evidence relationships: SQLite commits first, project backup artifacts refresh after commit, and sidecar/frontmatter compatibility output is refreshed only as generated transparency.

List asynchronous jobs currently known to this server. Use this when you lost a job_id or need a dashboard view of running/completed jobs. Returns an object envelope containing a jobs array of job objects sorted by newest first.

List canonical chapters for a project. Returns chapter_id plus compatibility sort order so callers can orient from numeric chapter aliases without treating those aliases as structural authority.

List indexed characters with their character_id, name, role, and arc_summary. Use this mainly as a lookup and disambiguation helper when you need to find a character_id for a broader reasoning task. Response shape note: returns a structured envelope (`results`, `total_count`).

List indexed places with their place_id and name. Use this mainly as a lookup and disambiguation helper when place context becomes relevant to the current reasoning task. Response shape note: returns a structured envelope (`results`, `total_count`).

List direct reference documents linked from a scene via metadata (for example, reference_ids). Returns only one-hop scene -> reference links and does not recursively traverse related references. If scene IDs are reused across projects, omitting project_id returns CONFLICT with candidate project_ids. Response shape note: returns a structured envelope (`results`, `total_count`) plus the resolved `scene_id` and `project_id` context.

List git commit history for a scene, with timestamps and commit messages. Use this to find commit hashes for get_scene_prose historical retrieval.

List subplot/storyline threads for a project. Returns a structured JSON envelope with results and total_count. Use this mainly as a lookup and disambiguation helper before deeper thread reasoning with get_thread_arc. Supports pagination via page/page_size.

Merge metadata directly from a Scrivener .scriv project into existing scene sidecars by starting a background job. This path is opt-in and requires sidecars to already exist (for example, from import_scrivener_sync). Returns immediately with a job_id to poll via get_async_job_status.

Move a scene through the explicit structure workflow. Writes canonical SQLite chapter linkage and/or timeline_position first, then mirrors compatibility fields to the scene sidecar and index; it does not move, rename, or resequence scene files or Scrivener-compatible folders.

Preview how explicit updates would change an existing prose-styleguide.config.yaml without writing any files.

Dry-run planning tool for review bundles. Resolves scene scope, deterministic ordering, warnings, and planned output filenames without writing files. Rendering options are accepted for API consistency and reflected in resolved_scope.options, but do not change planning output.

Generate a proposed revision for a scene. Returns a proposal_id and a diff preview. Nothing is written yet — you must call commit_edit to apply the change. This tool requires git to be available. If scene IDs are reused across projects, omitting project_id returns CONFLICT with candidate project_ids.

Record how two characters relate in a specific scene. This outcome-level workflow writes character relationship beats directly to SQLite, validates scene evidence, refreshes project backups after commit, and does not require callers to know the character_relationships table.

Rename a canonical chapter through the explicit structure workflow. Updates canonical chapter state and explicit scene chapter_title compatibility fields; it does not rename scene files, sidecars by path-derived structure, or Scrivener-compatible folders.

Reorder a canonical chapter through the explicit structure workflow. Updates canonical chapter order and explicit scene chapter/chapter_title compatibility fields; it does not rename, move, or resequence scene files, sidecars by path-derived structure, or Scrivener-compatible folders.

Explicitly restore a project from a trusted generated project backup bundle. Defaults to dry-run planning; dry_run=false applies canonical SQLite changes transactionally after the reviewed current snapshot checksum and required destructive or cross-scope confirmations are provided. Restore plan and confirmation-refusal responses include compact plan_summary details; confirmation refusals include blocking_requirements. include_unchanged=false suppresses unchanged row details while preserving summary counts.

Explicitly restore canonical SQLite chapter, scene-placement, and epigraph structure from a trusted generated structure export. This is a repair workflow, never a sync side effect; it validates project identity, schema, checksum, file presence, and conflicts before applying a transaction.

Keyword/FTS metadata search across scene titles, loglines (synopsis/logline text fields), and indexed metadata keywords (tags/characters/places/versions). Use this when you know likely words or exact metadata values but do not know the scene_id or chapter. This is not semantic search and does not search prose text; use find_scenes for structured filters and get_scene_prose after identifying likely scenes. Supports pagination via page/page_size and auto-paginates large result sets with total_count.

Full-text search across indexed reference document titles, summaries, and tags. Use this to discover world-building notes, continuity references, research docs, and other reference material without loading full file contents. Response shape note: returns a structured envelope (`results`, `total_count`).

Create prose-styleguide.config.yaml at sync root or project root using language defaults plus optional explicit overrides.

Generate skills/prose-styleguide/SKILL.md from the resolved prose styleguide config and universal craft rules. Optionally publish AI boot files (CLAUDE.md and .github/copilot-instructions.md) when using sync-root config scope.

Manually create a git commit (snapshot) for the current state of a scene. Use this to mark important editing checkpoints outside of the propose/commit workflow.

Suggest reference documents for a scene by aggregating links from the scene's characters and places. Returns weighted candidates ranked by how many entities in the scene link to each reference. Excludes any explicit scene reference links already present. In apply mode, selected links commit to SQLite first, project backups refresh, and scene sidecar/frontmatter output is refreshed only as generated compatibility.

Summarize the currently resolved prose styleguide config in plain language for review or confirmation.

Re-scan the sync folder and update derived scene/character/place indexes from disk. For already managed projects, sync reports file-derived chapter or epigraph drift without adopting it as canonical structure; use explicit import, repair, or structure tools for structural changes.

Track a storyline, subplot, or recurring arc through a scene by recording the scene's thread beat. This is the outcome-level workflow for thread relationship changes: callers provide story intent, while SQLite thread tables, backup refresh, and rollback stay implementation details.

Update canonical character profile fields such as name, role, arc_summary, first_appearance, and traits. SQLite commits first, project backups refresh after commit, and the .meta.yaml file is refreshed only as generated compatibility output; prose notes are never modified.

Update canonical place profile fields and retained compatibility notes. The place name commits to SQLite first and refreshes project backups; associated_characters and tags are compatibility/review metadata only. Use connect_character_place_evidence when scene-backed character/place evidence is paired, or connect_scene_place_evidence when scene evidence is place-only.

Update an existing prose-styleguide.config.yaml at sync-root or project-root scope by writing only explicit field changes.

Update one or more non-structural, non-relationship metadata fields for a scene. Writes only supplied allowed fields to the .meta.yaml sidecar and preserves existing structural compatibility fields; it never modifies prose, mirrors path-derived structure, or changes scene character/place relationship authority. Structural fields (part, chapter, chapter_id, chapter_title, timeline_position) are rejected here; use list_chapters plus assign_scene_to_chapter, move_scene, rename_chapter, or reorder_chapter for structure changes. Relationship fields (characters, places) are rejected here; use discov…

Compatibility name for link_reference_evidence. Prefer link_reference_evidence when recording story evidence; this retained alias still validates, commits SQLite first, refreshes project backups, and treats sidecar/frontmatter output as generated compatibility.

Compatibility name for track_thread_arc. Prefer track_thread_arc when recording story intent; this retained alias still validates, writes SQLite first, refreshes project backups, and rolls back failed canonical writes.