Skip to main content
The Manual

Documentation

Ash Tools & Skills

MCP tool calling, approval gates, the skills catalog, and persistent memory: how agentic Ash works, and what it will and won't do without you. Also covers connecting Claude Code, Claude Desktop, Cursor, ChatGPT, or any other MCP client directly to your account.

Ash has hands now

Beyond Generate PRO, Revise, and Human Grit Pass (see AI Generation), Ash can call real tools straight from chat: create or update a chapter, manage your story bible, run a self-check against known AI tells, or install a focused skill. Every manuscript write waits for your approval by default.

The Tool Catalog

These tools are always available in Ash chat, on top of the manuscript and story bible it already sees. Find them listed in the Ash Tools panel (the wrench icon in the chat header), grouped the same way here.

Manuscript

  • • Create a scene or chapter — add a new chapter or scene, ready to draft into
  • • Save into a scene or chapter — write or replace a chapter's content
  • • List chapters / Read a chapter — look up or fetch a chapter missing from the current briefing

Story bible

  • • Add to the story bible — create a character, location, faction, item, or lore entry
  • • Edit the story bible — update an existing world entry
  • • Remove from the story bible — delete a world entry, with confirmation
  • • List / Read a world entry — look up or fetch full detail on one entry

Prose

  • • Generate prose — draft prose for a chapter or scene, grounded in the story bible
  • • Revise text — rework a passage against an instruction, keeping the story's voice
  • • Human Grit Pass — roughen up overly polished AI prose
  • • Editorial Pass — structured line and structural notes on a chapter
  • • Slop self-check — score a passage for AI tells 0-100, with the worst offenders listed, drawing on the same corpus behind the public Slop Checker

Planning & skills

  • • Update task plan — a durable per-story to-do list, so multi-step work survives across turns
  • • List skills / Load a skill — check what's installed, and read a skill's instructions before using them

Memory

  • • Remember — save a durable preference, correction, or project fact that survives across chat sessions
  • • List memories / Forget — see or delete what's saved

Approval Gates

Off by default

By default, Ash can't create a chapter, save into one, or generate prose without asking first. Instead, it proposes a plan, and you see an approve or deny card right in the chat before anything is written.

The approval card names exactly what's about to happen: a new chapter or a specific existing one, and a "Will apply" line naming the Voiceprint, Story Rules, and canon world entries the draft will draw on. Approve it and Ash writes; deny it and nothing changes.

Trust the flow and want it faster? Toggle Auto-approve manuscript writes on in the Ash Tools panel, and Ash creates chapters and drafts prose immediately, no confirmation step. You can switch it back off at any time.

The Skills Catalog

Skills are prose instructions, not code, that shape how Ash uses the tools above. Install one from the Ash Tools panel without leaving the chat, and it applies starting your next turn. Every first-party skill declares explicit limits on what it will call and won't.

Editor-in-Chief

Line-level editing pass, trims flab, tightens dialogue tags, flags repeated sentence shapes. Revises only, never drafts or touches the story bible.

Continuity Keeper

Cross-references world entries against manuscript text and flags contradictions. Read-only, reports and stops rather than fixing anything itself.

The Tell

Rewrites AI-cliché prose using the same corpus behind the Slop Checker, with measured before/after scores. Revises existing prose only.

Voice Coach

Judges a chapter against the story's pinned voice print and proposes dial-anchored line fixes. Coaches and revises only.

Structural Pass

Beat-level pacing review, where a scene sags, where a reveal lands too early. Never edits prose directly.

Drafting Partner

The one skill that can draft new scenes outright. It flags missing canon instead of inventing world entries on its own.

Cold Reader

Reads as a first-time reader would, reporting confusion points without reading ahead or consulting the story bible.

The full catalog, including any custom skills across your API connections, lives in Settings → Skills.

Memory

Ash can save four kinds of memory, and re-reads all relevant ones on every turn:

  • • Preference — how you like things done
  • • Feedback — a correction you gave it
  • • Context — background worth remembering
  • • Reference — a fact to point back to later

Each memory is either global (applies everywhere) or scoped to a single story. Ash is instructed to trust what you've told it to remember over its own guesses, and to save a new memory when you explicitly ask, correct it, or confirm an unusual approach worked.

Where to find it

Open any story, then Ash chat. Click the wrench icon in the chat header for the Ash Tools panel: the auto-approve toggle, the built-in tool catalog, and the skills you can install or remove for that connection.

Connect Your Own AI Agent

Ash isn't the only way in

Every tool in the catalog above is also a real, remote MCP server, so Claude Code, Claude Desktop, Cursor, or any other MCP-aware tool on your own machine can work directly against your AIStoryHub account — your stories, story bible, prose tools, skills, and memory, from wherever you already work.

Claude Code & Cursor

Create a scoped API token in Settings → Developer API — pick a preset (full access, read-only, or prose-only) instead of hand-picking every scope — and copy the ready-made claude mcp add command shown right there with your token already filled in.

Claude Desktop & claude.ai Connectors

These support one-click "Add connector" against a plain MCP URL — no token to copy. AIStoryHub runs a real OAuth 2.1 authorization server behind it: connecting prompts you to sign in and approve the exact permissions requested, and the resulting connection shows up in Settings → Developer API right alongside manually-created tokens, revocable the same way.

ChatGPT

ChatGPT's Developer Mode (Settings → Apps → Advanced settings, on Pro, Plus, Business, Enterprise, and Education accounts) adds custom connectors over the same Streamable HTTP + OAuth 2.1 shape — point it at the MCP endpoint on /developers and OpenAI drives the same sign-in and approval flow.

Skills travel with you

Any skill installed on a connection — from the catalog above, or one you wrote yourself — can be downloaded as a real SKILL.md file from Settings → Skills, ready to drop into .claude/skills/ in any local project. It keeps working there even outside an active AIStoryHub story.

Full endpoint reference, scopes, and setup instructions for every client live at /developers.