MCP App Store
by KeenEthics
ArcanoLabs icon

ArcanoLabs MCP Server for ChatGPT

By ArcanoLabs

Overview

ArcanoLabs is a writing and publishing workspace, and this plugin is how an ArcanoLabs account is operated from inside ChatGPT. Its purpose is to carry a piece of writing from idea to published page: plan and draft an article or a book chapter, keep a long manuscript internally consistent as it grows, attach images to it from an existing library, check what readers are actually searching for, and send the finished piece to a WordPress site — without leaving the conversation. In practice it replaces switching between a manuscript editor, an image library, a keyword tool, a contact list and a WordPress admin screen, and it acts only on the data in the account that signed in. Books, in Scrybe. List a library, read a book's Story Bible and chapter structure, create and arrange chapters, work with fragments and templates, and keep characters, places and story entities consistent as a manuscript grows. Scrybe was previously called StoryForge; some tool names still carry the older "story" spelling for backward compatibility. Articles and publishing. Plan and draft blog posts, score a draft before it goes out, list unpublished work, reuse or create images from an existing library, and publish or schedule posts to a connected WordPress site. Search and contacts. SEO tools surface underserved searches and watch rankings for domains the account owns. CRM tools read and update the account's own contacts, companies, deals and pipelines. Who it is for: writers, solo founders and small teams who already keep their work in ArcanoLabs. Requirements and limits. An ArcanoLabs account is required; sign-in happens through ArcanoLabs and the tools only ever reach that account's own data. Generation tools draw on the account's credit balance, which is held and managed entirely inside ArcanoLabs. Publishing to WordPress, scheduling a run of posts, and deletes with no restore path always ask before they run. ArcanoLabs cannot receive file uploads through ChatGPT: to attach an image it returns a short-lived upload link to open in a browser.

How to connect

  1. ChatGPT: Open the ArcanoLabs listing in the ChatGPT app directory and choose Connect. ChatGPT listing ArcanoLabs exposes 147 tools here.

Tools

add_studio_library_voice

ChatGPT
Adds a voice found with search_studio_voice_library to this account at the provider, which is what makes it speakable. Free — it spends no credits. Ask the writer before calling this, even though it costs nothing: it changes their voice list at the provider and consumes one of a finite number of plan slots. It is reversible — deleting the voice at the provider frees the slot — but it is their account, not a scratch space. The `voiceId` in the response is the id to cast with, and it is not always the `libraryVoiceId` you asked for. Store what comes back. It still needs an authorization before cast_studio_character will take it, and an assistant cannot create one of those — the writer grants it in the studio. A refusal here is a real answer, not something to retry: "already added" and "no slots left" need different things from the person reading it.

append_slide

ChatGPT
Put one more slide into an existing deck. If the deck is on screen the audience sees it arrive at once, so this is the tool for answering a question from the room mid-talk. Leave out afterIndex to add it at the end. Free — no credits.

apply_console_preset

ChatGPT
Load a saved tuning onto a book’s console. This replaces what is there — save the current settings as their own preset first if they are worth keeping. Free — no credits.

approve_mini_lesson_plan

ChatGPT
Mark the plan as approved. Still free and still renders nothing — render_mini_lesson_video is what spends. Approve only what the user has actually read.

arrange_chapters

ChatGPT
Change the reading order of chapters. Like arrange_fragments, you may list only the chapters you are moving; the rest keep their places. Free — no credits.

arrange_fragments

ChatGPT
Change the order of fragments inside one chapter. You do NOT have to send the whole chapter: list only the fragments you are moving, and they are placed into the slots they currently occupy, in the order you list them. Everything you leave out keeps its place. Example: a chapter reading [A, B, C, D] with fragmentIds [D, B] becomes [A, D, C, B]. Free — no credits.

attach_story_image

ChatGPT
Attach an image to a character, location or object from a URL ArcanoLabs can fetch — an image hosted on the open web, or the user’s own site. PNG, JPG, WebP or GIF up to 8MB. This is NOT the tool for an image in this conversation: a chat sandbox path (/mnt/data/…) and an expired signed link both fail here. Use create_image_handoff for those. imageBase64 exists for a client that can genuinely produce it, and is capped at roughly 1.5MB — prefer a handoff link over spending a context window on base64. Free — no credits.

cast_studio_character

ChatGPT
Gives a character a voice for one production. Takes an authorizationId from list_studio_voice_authorizations — the voice identity comes from that record, so you never pass a raw voice id and cannot cast something unauthorized. Reversible: casting again replaces the profile. Free — casting spends nothing; performing the lines is what costs. If it refuses with NOT_AUTHORIZED_VOICE the voice has no record of what it may be used for, or that record was revoked or has expired — tell the writer to grant it in the studio rather than retrying.

check_credits

ChatGPT
Check ArcanoLabs credit balances and costs. Omit toolId to get the full MCP cost table plus current balances. With toolId, returns canUse + cost for that tool. image-generation costs 800 mana / 80 arcane per image.

check_image_handoff

ChatGPT
Whether the user has dropped their image yet. Returns pending, completed (with the image URL) or expired. Free — no credits.

check_mini_blog_status

ChatGPT
Whether the Mini Blog connection, narration and video pipeline are configured on this account. Call this before offering video, so a missing connection is said once rather than discovered as a failure. Reports no secrets. Free — no credits.

check_ranking_watch

ChatGPT
Properties whose Search Console rankings dropped against their own 28-day baseline, worst first, lined up with Google’s published Search Status incidents and with ordered recovery steps. Free — no credits.

check_scene_art_kit

ChatGPT
Every slot of one kit, its label, the reference shape for a mouth, and whether it is filled — the whole drop page in one read. Free. Use this to answer "did they finish?" rather than asking the writer. readiness.renderable is the honest answer to whether it can be shot yet, and it counts only the required slots: a kit missing optional brows is perfectly shootable, so do not tell somebody their character is unfinished when it is not. missingRequired is what to ask for; missingOptional is a later layer nobody is waiting on.

check_studio_providers

ChatGPT
Which voice providers this environment has configured and what each can do. The honest first call before any casting: it reports capability rather than claiming it, so you can tell the writer "no voice provider is configured" instead of failing at the point of casting. Free — no credits. Needs the Acting Studio switched on; without it every studio tool answers "not found".

check_wordpress_connection

ChatGPT
Whether this account can publish to its own WordPress site. This is what gates the Arcane unlock — but WordPress is OPTIONAL: someone with no site publishes inside ArcanoLabs instead and still earns the unlock, so never present a missing connection as a blocker. Set probe to ask WordPress whether the stored credentials still work. Credentials are never returned. Free — no credits.

connect_funnel_pipeline

ChatGPT
Make this funnel open a deal in a sales pipeline for every lead it captures. Take pipelineId and the stage id from get_crm_pipeline_board — the stage is checked against that pipeline before anything is written, because an unknown stage would put every captured deal in a column no board renders, which the owner experiences as leads going missing. Both halves are required: a pipeline without a stage is not a destination. Free — no credits.

convert_prose_to_lines

ChatGPT
Reads a book's narrative prose and lifts the dialogue out of it as screenplay lines, so the scene can be cast. Cost: 100 mana / 10 arcane per passage — call it with dryRun first and you get the passage count and the price before anything is spent, which is free. Charged on delivery and per passage: a passage nothing could be lifted from costs nothing, and running out of credits part-way ends the run and keeps everything already converted rather than rolling it back. Spoken words are taken verbatim from the prose and a line the model could not match to the text is dropped, so this never puts words in a character's mouth that the author did not write. A fragment that already has lines is skipped, and a fragment that already reads as a screenplay belongs to set_down_screenplay, which is free. Needs the Acting Studio switched on.

create_automation

ChatGPT
Save a rule. It is stored SWITCHED OFF whatever you send — arming it is enable_automation, a separate decision, because a rule that fires writes to the owner’s real CRM on real leads. Conditions are ANDed; there is no OR and no branching, so two rules are clearer than one with a fork in it. Actions run in order and do not chain: nothing a rule does fires another rule. Call get_automation_vocabulary first for the exact event and field names. Free — no credits.

create_book_template

ChatGPT
Make a book template shaped like the book THIS writer is actually writing, and save it to their account. The shipped templates cover twenty common shapes; they do not cover a devotional whose chapters are a reading, a reflection and a prayer, or a repair manual that goes symptom, diagnosis, parts, procedure. You have just heard them describe their book, so you are the one who can write that shape down. A template is an opening chapter, one or more body shapes that cycle through the middle, and a closing chapter — each one a list of fragment beats in order. Call list_fragment_templates first: every templateId must be a real fragment, and a template naming one that does not exist is refused rather than saved. Call list_story_presets too and read a shipped one with get_story_preset if you want a worked example of the shape. Once saved it is a first-class template: it appears in the writer’s picker inside Scrybe, get_story_preset lays it out over any chapter count, and map_chapter_fragments takes its id as preset. Pass templateId to rewrite one that already exists. Free — no credits, because arranging a book of your own should cost nothing.

create_chapters

ChatGPT
Create chapters, in order, in one call — a whole outline goes in as one list of titles rather than one call per chapter. The chapters are created empty; no prose is written and nothing is charged. They go at the end of the book unless you pass position, and the book is renumbered either way. Add the beats inside them with create_fragments. Free — no credits.

create_crm_record

ChatGPT
File a new contact, company, deal, activity, note or pipeline. The fields each kind reads: contact firstName, lastName, email, phone, jobTitle, companyId, status, lifecycleStage, leadScore, tags, customFields — and it needs a name or an email, because a record that says who it is not is worse than no record; company name (required), domain, industry, companySize, website, phone, tags, customFields; deal name (required), amount, currency, stage, pipelineId, contactId, companyId, probability, expectedCloseDate, source; activity subject (required), type, description, contactId, companyId, dealId, dueDate, status; note content (required, and it has to be content — no other key is read), contactId, companyId, dealId, isPrivate; pipeline name (required), stages, isDefault. Linking by id updates the other record too, so a deal filed against a contact shows on that contact. Free — no credits.

create_deck

ChatGPT
Write a whole presentation to ArcanoLabs, ready for Scrybe to show on a television. Call get_deck_format first. Free — you wrote the deck, nothing of ours generates it.

create_fragments

ChatGPT
Create fragments inside one chapter — the beats the writer will fill in the studio. Leave text out and the fragment is an empty slot of that template, ready for the studio’s generation (which is where the writer spends credits and where the console controls apply). Pass text only for prose the writer already has: it is stored exactly as given, generated by nothing. Call list_fragment_templates first for a real templateId and its params. Beats meant to read as one continuous passage should share a groupId as you lay them out — see group_fragments for what combining does to the writing. Free — no credits.

create_funnel

ChatGPT
Build a new funnel. It is created as a draft and no page goes live — call publish_funnel when the owner has read it. One field must be type "email", or there is nothing to create a contact from. Point it at a pipeline with connect_funnel_pipeline so captured leads also open a deal. An email sequence cannot be set here and a "sequence" key is ignored, because scheduling mail to real people belongs behind a screen where the owner can read it first. The page's look is yours to set too: "brand" for the name, logo and colours, "layout" for how it is arranged, "sections" for the benefit blocks, and offer.imageUrl for the featured image. Free — no credits.

create_image_handoff

ChatGPT
THE WAY TO ATTACH AN IMAGE THE USER ALREADY HAS. Returns a short-lived link bound to one character, location or object. Give the user the link; they open it, drop the file, and it lands on the entity. Nothing goes through this conversation. Use this whenever the image exists as a picture rather than a public URL — an image generated in this chat, a file the user attached, a screenshot, anything on their phone. You cannot read those bytes and must not try: a generated image here lives at a sandbox path ArcanoLabs cannot fetch, and retyping an attachment as base64 costs a fortune in tokens and usually corrupts the file. The link lasts 30 minutes. Poll check_image_handoff to see when it has been used. Free — no credits.

create_story_book

ChatGPT
Create an empty book on this account and return its bookId. Use it when the writer names a book that list_story_books does not show. The book starts with no bible, no chapters and no prose: fill the bible with update_story_bible and upsert_story_entity, then add chapters with create_chapters. A title the account already uses is refused so a retry cannot leave two books with the same name — pass allowDuplicateTitle only if the writer really wants a second one. Pass scriptFormat when the writer is writing a film or a play rather than a book: screenplay gives slug lines, CUT TO: and V.O.; stage gives ACT ONE, Blackout and aside. Omit it for a book. Free — no credits.

delete_automation

ChatGPT
Remove the rule for good. There is no trash and no undo: the trigger, every condition and every action are gone. If the rule is misbehaving, disable_automation keeps it and stops it, which is almost always what is wanted. Free — no credits.

delete_book_template

ChatGPT
Remove one of the writer’s own templates. The templates ArcanoLabs ships cannot be deleted, and no book already laid out from this template is touched — the chapters and fragments stay exactly as they are. Free — no credits.

delete_console_preset

ChatGPT
Remove a saved tuning. Books already using it keep their settings. Free — no credits.

delete_crm_record

ChatGPT
Remove a contact, company, deal, activity, note or pipeline. There is no undo and no trash — the record is gone, and the history filed against it is orphaned rather than removed. Ask the account holder before every one of these, and prefer setting a contact’s status to lost or unqualified over deleting a person who may call back. Free — no credits.

delete_fragments

ChatGPT
Take one or more fragments out of a chapter for good. The studio asks the writer "this cannot be undone" before it deletes; you have no such moment, so confirm must be true and you only pass it once the writer has actually asked for the delete — not to tidy up on your own initiative. The fragments are kept in the chapter’s deleted list, so restore_fragments can put them back; the twenty most recent are kept. To take a fragment out of the AI’s context without losing it, use update_fragment with enabled: false instead. Free — no credits.

delete_image

ChatGPT
Delete one or more owned library images by storage path. Community images cannot be deleted.

disable_automation

ChatGPT
Stop the rule firing, keeping it and its run log. This is how you pause an automation that is doing the wrong thing — it is reversible, and it is always the right first move. Free — no credits.

discard_mini_lesson_plan

ChatGPT
Discard the plan so the next render plans fresh. The lesson and its video are untouched. Free — no credits.

draft_funnel_copy

ChatGPT
Draft a funnel’s headline, subheadline, form fields and success message, and/or its follow-up emails, in the account’s own brand voice. Pass brandId or the copy comes back in neutral prose and says so. It returns a DRAFT and saves nothing — apply it with update_funnel and set_funnel_sequence after the owner has read it. It will not write a URL anywhere: any link the model produces is replaced with "[your link]" and reported in warnings, because a made-up link in a real customer’s inbox is indistinguishable from a scam under the owner’s own name. It also invents no prices, client names or statistics — it leaves [bracketed] instructions instead, and those must be replaced before anything is sent. Pass the warnings on. Costs 300 mana (30 arcane), charged only when copy comes back.

enable_automation

ChatGPT
Arm the rule. From here on every matching event writes to the owner’s real CRM — tags change, tasks appear, deals open — on real leads, with nobody watching. Run test_automation first. Free — no credits.

enable_funnel_sequence

ChatGPT
Turn the sequence on. From here every person who submits this funnel is queued for these emails, where their consent allows it — real mail, to real addresses, while nobody is watching. preview_funnel_sequence first, and read the copy: a {{token}} typo or a placeholder left in a body is delivered exactly as written. Nobody who submitted earlier is affected. Free — no credits.

export_actor_packet

ChatGPT
The sheet a performer is handed. full (the default) prints whole scenes with their lines marked, because a column of sentences with no situation is the sheet nobody can act from; cues prints the line before each of theirs then theirs, for rehearsal; lines prints their dialogue only, for memorising. A writer-only note line never prints, in any mode. Nothing is stored — the packet is built from the book every time, so a correction made in a chapter is in the next packet with nobody rebuilding anything. That is also why version and draftDate are passed in and printed on every page: a sheet in somebody’s hands is a snapshot and has to say which one. Page numbers are an estimate, 55 lines a page, but a stable one, because a performer writes on them. Free — no credits.

export_crm_records

ChatGPT
Every contact, company or deal as CSV text, with a header row — the same export the app offers, returned as a string rather than a file so you can read it, summarise it or hand it on. Up to 5,000 rows unless you ask for more, 10,000 at most. Use this rather than paging list_crm_records when the question is about the whole set. Free — no credits.

generate_blog_post

ChatGPT
Write a complete HTML blog post. Cost: 4,000 mana / 400 arcane. Pass images as Firebase download URLs or library storage paths (resolved server-side). Use images: [] if none. Set dryRun to price without writing.

generate_image_prompt

ChatGPT
Craft a professional DALL·E image prompt for an article. Cost: 4,000 mana / 400 arcane. Set dryRun to validate without charging.

generate_images

ChatGPT
Generates one DALL·E image per outline point. Cost: 800 mana / 80 arcane per image. Returns per-item status (ok/failed) with library path; only successful images are charged. Set dryRun to price without generating.

generate_mini_lessons

ChatGPT
Generate complete Mini Lessons from the account’s brand context — the whole lesson on the card, plus an Explore Deeper drawer — and draft them to the connected WordPress site. Cost: 900 mana / 90 arcane per lesson; a failed generation is refunded. Lessons are created as DRAFTS for the user to review; never generate a batch without asking first.

generate_outline

ChatGPT
Produce a 4–6 point actionable outline for a topic. Cost: 4,000 mana / 400 arcane. Set dryRun to validate without charging.

generate_story_image

ChatGPT
Have ArcanoLabs generate an image from a prompt and attach it to a character, location or object. Cost: 800 mana / 80 arcane per image — the same as the app charges for one image; a failed generation is not charged, and dryRun prices it without generating. Use this when the user wants a new image made. When they already have one, use create_image_handoff instead — that is free. Write the prompt from the entity’s own description in the story bible so the portrait matches the book.

generate_voice_sample

ChatGPT
Write a short sample post in a DRAFT brand voice so the user can answer "does this sound like you?" before anything is saved. Nothing is persisted. Cost: 100 mana / 10 arcane; a failed generation is not charged, and dryRun prices it without generating. Supply whichever you have: the user's own writingSample (best — a sample beats a questionnaire), a voiceProfile, or a draft brandData with a tone.

get_article_plan

ChatGPT
Everything the ArcanoLabs writer is told before it writes a post on this topic: the brand dossier, the closed set of URLs the article may link to (ranked, the brand's own page first), the competitors it must not recommend, the closing call to action, the queries Search Console says the site fails to serve, and what the published-article grade will measure. Write the draft against this, then call score_article_draft. Free — no credits.

get_automation

ChatGPT
One rule in full: its trigger, every condition and every action. Read this before updating one, because an update replaces the conditions and actions you send. Free — no credits.

get_automation_runs

ChatGPT
What this rule actually did, newest first: which event fired it, which actions ran, and the error on any that failed. This is the one place that answers "why did my rule not do anything?" — a rule with no runs never matched, and a rule with failed runs matched and could not act. Free — no credits.

get_automation_vocabulary

ChatGPT
The events a rule can fire on, every field a condition can read, the comparisons available, and the actions a rule can take — read this before writing a rule rather than guessing, because a field name that is not on the list is refused on save. Free — no credits.

get_brand_profile

ChatGPT
The resolved Brand Operating Profile for one brand: the voice, audience, themes, prohibitions and absolute rules every ArcanoLabs generator writes against. Read it before drafting anything in the user’s voice, and honour its prohibitions. Composed from existing records — free, no credits. Use get_brand_wizard first for the list of brands and their ids.

get_brand_wizard

ChatGPT
List the brands this ArcanoLabs account already has, with whether each carries a voice profile. Call this FIRST during onboarding: if a brand already exists, ask whether to update or add another rather than creating a duplicate. Free — no credits.

get_chapter_fragments

ChatGPT
The fragments of one chapter in order, each with its position, template, status, groupId, word count and a short preview. Read this before arranging anything, so you move the fragment the writer means. Fragments sharing a groupId are combined into one passage and written as one; group_fragments combines and separates them. includeText returns the full prose — large, so leave it off unless you are actually editing. Free — no credits.

get_console

ChatGPT
The saved plugin console for one book: the class racks (tension, pacing, subtext, naturalism and the rest), the style and writing-type controls, the per-character, per-location and per-event racks, the song rack and the modulation routes, plus the preset it came from if any. This is what the studio loads and what generation reads. Read it before tune_console when you are changing a list the whole of which gets replaced. Free — no credits.

get_console_controls

ChatGPT
Every control the ArcanoScript console exposes, with its type and allowed values: the per-class racks, the style and writing-type racks, the per-character, per-location and per-event racks, and the song rack. Call this before tune_console or before setting controlOverrides — a control that is not on this list is dropped on save, so guessing wastes a turn. Numeric controls are 0-100. Free — no credits.

get_console_preset

ChatGPT
One preset in full, including every control it sets. Free — no credits.

get_crm_contact_timeline

ChatGPT
Everything filed against one contact — calls, emails, meetings, tasks and notes logged as activities — newest first. This is what to read before writing to someone, so the next thing said to them follows the last. A contact belonging to another account reads as not found rather than as an empty timeline. Free — no credits.

get_crm_pipeline_board

ChatGPT
A pipeline with its stages and, with stats, the count and value of the deals sitting in each one — the Kanban board as numbers. Leave pipelineId out for the account’s default pipeline, which is where deals land when nothing says otherwise; an account with no default gets null rather than an error. Free — no credits.

get_crm_record

ChatGPT
One contact, company, deal, activity, note or pipeline in full, including its custom fields and the ids of what it is linked to. A record belonging to another account reads as not found, which is the right answer to both questions. Free — no credits.

get_deck

ChatGPT
One deck with its slides, in order. Read it before appending if you need a slide index — order is renumbered on every insert, so an index from an earlier call may be stale. Free — no credits.

get_deck_format

ChatGPT
Every deck kind, layout, theme, stage size and cap, with the rules, straight from the server that normalizes decks. Call this before writing your first deck. Free — no credits.

get_fragment_history

ChatGPT
The earlier versions of one fragment, newest first, alongside what it says now. The studio keeps the last ten; version 1 is what revert_fragments would put back. Read this before offering to undo a rewrite, so the writer is told what they would get rather than asked to take it on trust. A version longer than 20,000 characters comes back cut, marked truncated. Free — no credits.

get_funnel

ChatGPT
One funnel in full: its fields, offer copy, success message, the pipeline and stage its leads open a deal in, and its public URL if it is live. Read this before updating one, because an update replaces the fields you send. A funnel belonging to another account reads as not found. Free — no credits.

get_funnel_report

ChatGPT
One funnel’s numbers over a window: page views, form fills, contacts, deals opened, deals won, the value won, and what the follow-up email did — queued, sent, suppressed, failed. Email opens are not reported and never will be: mail clients prefetch tracking pixels, so an "open" describes a client rather than a reader, and reporting one as engagement would be telling the owner something false about their copy. A view is a page LOAD, not a person — obvious crawlers are dropped but nothing is deduplicated per visitor, so quote the rate as form fills per page view and never as a share of people. The response carries these caveats; pass them on rather than presenting the figures bare. Free — no credits.

get_funnel_scheduled_sends

ChatGPT
The queue for one funnel: who, which step, when, and what happened — pending, sent, suppressed (their consent list stopped it), failed or cancelled. This is how you confirm before an event that the reminders exist, rather than finding out on the day. No opens and no clicks, here or anywhere: a prefetched pixel describes a mail client, not a reader. Free — no credits.

get_funnel_sequence

ChatGPT
The saved sequence for one funnel: every step with its anchor, offset and copy, whether it is switched on, and whether the funnel confirms addresses by email first. Read this before editing — set_funnel_sequence replaces the whole thing. Free — no credits.

get_funnel_submissions

ChatGPT
What people actually typed into one funnel, newest first, with the attribution captured at the time — UTMs, referrer and landing page. These are real contact details, so treat them as the owner's customer data and not as material to repeat elsewhere. The contact records themselves are in the CRM: use get_crm_contact_timeline to see what has happened since. Free — no credits.

get_generation_status

ChatGPT
Whether a running blog, video or workflow generation has finished. complete means SUCCEEDED, not merely stopped — check failed separately. Drives the wait-screen cut, so poll this rather than assuming a fixed duration. Free — no credits.

get_image_url

ChatGPT
Mint a short-lived signed URL for a library storage path. Default TTL 3600s, max 86400s.

get_mini_blog_schedule

ChatGPT
Whether lessons are generated on a schedule for a brand, how many a week and at what hour. Free — no credits.

get_mini_lesson

ChatGPT
One lesson in full: its content, status, hero image, video state and WordPress link if it has one. A lesson that lives only on the platform has no WordPress page, and this is how its owner reads it. Free — no credits.

get_mini_lesson_plan

ChatGPT
The beats a Mini Lesson video would be rendered from, with the script controls behind them (target duration, speaking rate, reading level, energy, person, words to avoid, how hard the call to action pushes) and whether the plan is approved. Free — no credits.

get_post

ChatGPT
One post in full, body included, so it can be revised in this conversation. Pair it with publish_to_wordpress to push the revision back. Free — no credits.

get_post_performance

ChatGPT
A published post's own view counts plus the keyword positions this user tracks. hasRankingMovement true is the strongest moment in the funnel — surface it immediately. Free — no credits.

get_story_bible

ChatGPT
A book's Story Bible: premise, plot, genre, mood, themes, style, and its characters, locations, objects, events, methods, motivations and plot points. Read this BEFORE adding anything, so you update the character who already exists instead of creating a second one. Narrow with kinds when you only need part of it. The ids it returns are what tune_console takes for the per-character, per-location and per-event racks. Free — no credits.

get_story_preset

ChatGPT
One template, laid out over a book of this length: chapter by chapter, the shape each chapter takes and the beats inside it, each with a note saying what it is for. Read this before mapping a book if you want to show the writer the plan first, or adjust it. The opening and the closing are their own shapes and the middle cycles, so a long book is not the same chapter thirty times. Free — no credits.

get_story_stats

ChatGPT
Chapter count, word count and a per-chapter breakdown for one book. Free — no credits.

get_studio_project

ChatGPT
The production and every scene in it, in story order, with the real scene ids — which is what prepare_studio_scene needs. Scene NUMBERS move when scenes move, so always carry the id. Free — no credits.

group_fragments

ChatGPT
Combine consecutive fragments into a run that is written as one continuous passage, or pass groupId: null to break a run apart. This changes the writing, not just the layout: each fragment in a run is generated knowing the others, so a Monologue combined with a Setting Description and a Cliffhanger comes out as one scene that flows, instead of three passages stacked up. Uncombined, each is written as if it stood alone. Reach for it whenever consecutive beats are meant to read as one moment — a scene and the line that lands it, a description and the thought it provokes — and leave beats that are separate moments uncombined. Only fragments that already sit next to each other can be combined; arrange_fragments first if they do not. read_book and get_chapter_fragments show each fragment’s groupId, so you can see what is already combined. Free — no credits.

import_manuscript

ChatGPT
Takes a book the author has ALREADY written and puts it into a Scrybe book, cut into the chapters it already has. Pass the manuscript itself as text — an MCP client cannot hand over a file, so paste or read the manuscript in and send it. Chapter breaks are found from the lines that name a chapter ('Chapter 4', 'Kapitel 4', 'Capítulo 4', '第四章', 'Глава 4', a bare number), from Markdown '#' headings, or from the blank-line gaps between chapters. ALWAYS call with dryRun first and show the author the chapter list before writing: if no chapter break is found the whole manuscript becomes one chapter, which is almost never what they wanted, and the result says so rather than pretending it worked. Importing a .docx keeps the Word heading styles and is more reliable — that path is the Import button in Scrybe. Rewrites nothing and invents nothing: it is the author's own words, cut where the author already cut them. Calls no model, so free — no credits.

list_automations

ChatGPT
Every automation rule on this account: what it fires on, whether it is switched on, how many times it has run and what failed last. An automation is "when this happens in the CRM, if these conditions hold, do this" — tag the contact, move its lifecycle stage, file a task, open a deal, or email the account owner. Free — no credits.

list_chapters

ChatGPT
Chapters of one book in reading order, with fragment counts and word counts. Start here before moving anything — every fragment tool needs a chapterId. Free — no credits.

list_character_lines

ChatGPT
The character page. Every scene they are in — present and silent counts, not only speaking — with the scene number, slug line, status, who else is there, and the goal, emotion and knowledge the scene notes for them. Then every line they speak in the whole book in story order, each carrying what the scene says they know at that moment, which is how you catch a line that knows something the character has not learned yet. A character who speaks in a scene they were never listed on is in it too — the lines are the harder evidence. Reads the chapters and the story bible together, so the character has to exist in the bible. Free — no credits.

list_console_presets

ChatGPT
The console presets on this account — saved tunings that can be dropped onto any book. Free — no credits.

list_crm_records

ChatGPT
Contacts, companies, deals, activities, notes or pipelines for this account, newest first. Filters apply to the kinds that have them: contacts take status, lifecycleStage, companyId, ownerId and tags; companies take industry, companySize, ownerId and tags; deals take pipelineId, stage, contactId, companyId and ownerId; activities take type, status, contactId, companyId, dealId and ownerId; notes take contactId, companyId, dealId and ownerId. Pipelines take no filters and come back whole, with their stages. 50 records unless you ask for more, 200 at most — this is a list, not an export; use export_crm_records for the whole set. Free — no credits.

list_decks

ChatGPT
The decks on this account, newest first. Titles and ids only — the slides are not included, so use get_deck to read a deck's contents. Free — no credits.

list_fragment_templates

ChatGPT
The fragment templates the studio knows, with the params each one takes. Call this before create_fragments: templateId must be one of these ids, and a param you leave out survives into the fragment's prompt as a literal placeholder. Free — no credits.

list_funnels

ChatGPT
Every funnel on this account: its name, status, public URL once published, how many submissions it has taken, and whether an email sequence is attached. A funnel is a hosted landing page with a form that files whoever fills it straight into the CRM — it is not a sales pipeline, which is the stage list a deal moves along. Free — no credits.

list_images

ChatGPT
List images in the caller's ArcanoLabs library. Default scope is mine (private+shared). Returns path (not permanent URLs), pagination cursor, and total. Use get_image_url for a short-lived URL.

list_mini_blog_voices

ChatGPT
The narration voices available for Mini Lesson video, and whether the Mini Blog connection is configured. Call before asking the user to pick one. Free — no credits.

list_mini_lessons

ChatGPT
Mini Lessons this account has generated, with their status and WordPress links. Free — no credits.

list_post_plans

ChatGPT
Plans and their scores. Filter by status — draft (not approved yet), approved (ready to schedule), scheduled (on the calendar). Free.

list_posts

ChatGPT
The posts on this ArcanoLabs account — drafts and published alike — newest first, with ids, titles and status. Use it to find a post before reading or revising it. Free — no credits.

list_scene_art_kits

ChatGPT
The kits this account has, narrowed by book or by kind. Free. Use it to find a kit again in a later conversation instead of making a new one — and remember art is deliberately NOT tied to a scene: one drawing of a room serves every scene set there, so a kit found here is the thing to reuse rather than redraw.

list_scenes

ChatGPT
Every scene in reading order, with its number, slug line (INT./EXT., where, time of day), title, who is present, what turns in it, its status and the fragments in it. Pass a bookId for the whole book or a chapterId for one chapter — one or the other. character narrows it to the scenes someone is in, and presence counts, not only speaking: a character listed on a silent scene is in that scene. Scene numbers come from the order and are not stored, so they move when scenes move — quote a sceneId, not a number, when you want to come back to one. Free — no credits.

list_search_opportunities

ChatGPT
Scored page × query rows from the user’s own Google Search Console: searches their site nearly answers but does not. The strongest input to "what should I write next" — far better than guessing a topic. Reads the last scan rather than running one, so it is free. An empty result means Search Console is not connected yet, or nothing has been scanned.

list_sequence_templates

ChatGPT
The shapes a nurture sequence can take, with the day each email lands on and its subject line. Three of them: a five-step service follow-up (deliver, teach, prove, ask, leave the door open), live-session reminders anchored to a start time, and a three-email guide download. Pass an id to get one back in full, ready to save with set_funnel_sequence. The bodies come back with [bracketed] lines still in them on purpose — they are a shape and a first draft, not finished copy, and no template contains a URL, because a made-up link in somebody’s real email is the worst thing this can produce. Free — no credits.

list_story_books

ChatGPT
List the books on this ArcanoLabs account, newest first, with their ids. Call this before any other story tool — every one of them needs a bookId. Free — no credits.

list_story_presets

ChatGPT
Every book template this writer can use: the ones ArcanoLabs ships — epic fantasy, thriller, memoir, tool manual, handbook, show episode and the rest — and the ones they have made for themselves, marked source "custom". Each says what kind of book it is for and how many chapters it suggests. Pass an id as preset on map_chapter_fragments and every chapter takes that template’s shape, so a beginner never has to work out which fragments a chapter needs. Read this before writing a new template, so you extend what fits rather than duplicating it. Free — no credits.

list_studio_voice_authorizations

ChatGPT
Each record says which voice may be used for what — playback, conversion, synthesis, reuse, training, commercial — and whether it has been revoked or has expired. This is the list to cast from: cast_studio_character takes an authorizationId, not a voice id, so a voice nobody authorized cannot be cast by any route. An assistant cannot create one of these; it is the consent record for somebody's voice and the writer grants it themselves in the studio. Free — no credits.

list_studio_voices

ChatGPT
The voices on this account at the provider, each with a previewUrl you can hand the writer to listen to — the provider's own hosted sample, so nothing is generated and nothing is charged. Having a voice here is NOT enough to cast it: casting takes an authorization id, so check list_studio_voice_authorizations for what a voice is actually cleared for. Free — no credits.

list_unpublished_posts

ChatGPT
Posts that were generated but never went live, from both the schedule and the blog library. Use it to offer "want me to push this live?". A post marked failedPreviously errored on a previous publish and is invisible to the user unless you surface it. Free — no credits.

map_chapter_fragments

ChatGPT
Lay out a whole book in one call — the chapters, and the fragments each one is made of. This is the fast path after a story bible: instead of create_chapters then one create_fragments per chapter, send the map and it is applied in order. A chapter already in the book is matched by chapterId, or failing that by title, and topped up rather than duplicated, so re-sending a map after an error is safe, and a chapter that already has fragments is left alone when its fragments would have come from a template. Nothing is ever deleted. Either list each chapter’s fragments (templateId from list_fragment_templates) or name a template from list_story_presets; a non-fiction template also sets the book to write like a manual rather than a novel. Pass dryRun to see exactly what would happen without writing. The fragments arrive empty, ready for the studio’s generation. Free — no credits.

move_crm_deal

ChatGPT
Drag one deal to a different stage of its pipeline, in one call. Use this rather than update_crm_record for a stage change: it is the path the board itself uses, so the deal and the pipeline stay in step. Take stageId from get_crm_pipeline_board — a stage name is not a stage id. Free — no credits.

move_fragment

ChatGPT
Take one fragment out of its chapter and put it into another at a chosen position, or move it within its own chapter by omitting toChapterId. Both chapters must be in the same book. The fragment leaves any group it was in, because a group only means something between neighbours. Free — no credits.

open_scene_art_kit

ChatGPT
The way art you generated gets into a production. A kit is the set of drawings one character or one location needs, as parts that have to agree with each other: a character is thirteen required slots (head, body, two eye states and the nine mouth shapes), a location is one plate plus optional lighting states. Free. This returns the kit, every slot with its label, how far along it is, and a drop link. You generate the art and hand the writer the link; they drop the files on that page and the parts land in the library. No image bytes go through this tool — do not try to send one. Call it with the same `subject` twice and you get the same kit back, not a second one, which is deliberate and important: two kits for "Mark" means his thirteen drawings land split across both and NEITHER can ever be rendered, with nothing failing at the time it happens. Read created to tell the writer which happened — "I have made Mark a kit" and "Mark already had one, here is where to drop" are different sentences and saying the wrong one is confusing. The link asks only for the slots still empty, so calling this again for a half-filled kit is the right way to get somebody back to the parts they have not done. A kit with nothing left returns handoff: null — that is "it is finished", not a failure. For a character, the parts normally come out of ONE reference sheet you generate, drawn in a grid with the nine mouths on it — so generate the sheet, then tell the writer which slot each piece goes on. Each mouth slot carries a hint describing the shape it wants (mouth-A is lips together, M/B/P); pass those on, because nothing can check that the file dropped on `mouth-A` holds the A shape and a wrong one produces a face that moves wrongly rather than an error.

open_studio_project

ChatGPT
Lays a Scrybe book out as a production: its screenplay lines become scenes with cast attached. Idempotent, so this is also how you find an existing production — calling it for a book that already has one re-reads it and reports what changed instead of making a second. Use the project id it returns for get_studio_project. A book whose chapters are still prose produces a production with no dialogue: run convert_prose_to_lines or set_down_screenplay first, then open it again. Free — reads chapters that are already written and calls no provider.

plan_mini_lesson_video

ChatGPT
Work out the beats, narration and runtime for a lesson’s video before anything is rendered. Free: this plans only, and the render credit is charged later by render_mini_lesson_video. Planning again replaces the plan and returns it to draft. Controls persist across re-plans.

plan_posts

ChatGPT
Turn a list of headlines into one scored plan each, BEFORE anything is written. Each plan carries what the writer will be told: the brand, the only URLs the post may link to (ranked, the brand's own page first), the competitors it must not recommend, the closing call to action, and the searches this site shows for and does not answer. The score says whether the post is set up to succeed and names what is missing. Nothing is generated, charged or published. Free. At most 50 headlines per call.

prepare_studio_scene

ChatGPT
One scene in full: its lines in order, its cast resolved, and uncast — the characters with no voice yet, which is the list to work through before anything can be performed. It refuses a scene with no dialogue rather than answering with an empty scene, and that refusal is information: the chapter is still prose, so run convert_prose_to_lines or set_down_screenplay and open the production again. Free — no credits.

preview_funnel_sequence

ChatGPT
What would be queued for somebody signing up right now, with the real send times, computed by the engine that actually sends them. Also names the steps the planner would DROP and why — an event reminder whose moment has passed is dropped on purpose, and a preview that hid that would teach the wrong thing. Pass a sequence to try an unsaved one. Sends nothing, saves nothing, queues nothing. Free — no credits.

publish_funnel

ChatGPT
Put the funnel's page on the open internet under the owner's brand, at a URL that can be pasted into an ad or an email. Ask first, every time: this is visible to anyone with the link the moment it returns, and a URL that has gone out cannot be recalled. Read the funnel back with get_funnel and let the owner see the copy before you call this. Returns the live URL. Free — no credits.

publish_mini_lesson_to_archive

ChatGPT
Opt a lesson into the public /mini-blogs archive, or take it back out. Generation always leaves a lesson as a draft and nothing reaches the archive until this is called, so ask the user first — this is the step that makes their writing public. Free — no credits.

publish_mini_lesson_to_youtube

ChatGPT
Upload a rendered Mini Lesson video to the user’s own YouTube channel, point the lesson card at the YouTube URL and delete the copy we were hosting. Ask the user every time: it posts to their channel and the original file does not come back. Default visibility is unlisted. Free — no credits.

publish_to_wordpress

ChatGPT
Publish / post a blog article to WordPress (WP publish). Cost: free. Creates or updates a WordPress post on the connected site. Set dryRun to validate without publishing.

read_book

ChatGPT
The book as a reader meets it: chapters in order, fragments in order, with their text. This is the tool for being asked to read a book and say what is wrong with it — get_chapter_fragments shows one chapter and previews rather than prose. A long book comes back a few chapters at a time: the result carries nextChapterId when there is more, and it always stops on a chapter boundary so nothing is judged on a sentence that was cut. Empty fragments are left out unless you ask for them. Consecutive fragments sharing a groupId are one combined passage, written as a single run; where a stretch reads as separate blocks that should have been one moment, group_fragments is the fix. Free — no credits.

regenerate_mini_lesson_cover

ChatGPT
Rebuild the designed cover for a lesson and, when it is already on YouTube, push it as the custom thumbnail. This is for lessons that shipped before covers were designed and carry an auto-picked frame instead.

remove_story_entity

ChatGPT
Remove one entity and add its name to the story bible’s excluded list, so later generation stops referencing it — a cut character otherwise keeps walking into new chapters. Confirm with the user before calling this: it is their book. Free — no credits.

render_mini_lesson_video

ChatGPT
Render the video for a published lesson. Costs one video credit, charged only after the content and a ready template both check out — a catalog or contract miss is reported and charged nothing. Returns as soon as rendering starts; read the lesson back with get_mini_lesson for progress, and the MP4 attaches itself when it is done. Plan it first if the user wants a say in the beats.

restore_fragments

ChatGPT
Put deleted fragments back into their chapter, with their text and params as they were. Omit fragmentIds to restore everything in the chapter’s deleted list. They come back at the end of the chapter unless you give a position, because where a fragment sat is not where it belongs once the chapter has moved on — arrange_fragments puts it right. Free — no credits.

revert_fragments

ChatGPT
Put fragments back to the version before the last rewrite — one step back per call, off the same history the studio keeps. Omit fragmentIds to take the whole chapter back a step. A fragment with no earlier version is left as it is and named in the result rather than being emptied. Call get_fragment_history first when the writer should see what they are going back to. Free — no credits.

rewrite_fragments

ChatGPT
Replace the text of many fragments of one chapter in a single call — the tool for "go through this chapter and tighten it". One fragment on its own is update_fragment. Each fragment keeps its previous version, so revert_fragments can put the chapter back, but it is still an overwrite: confirm must be true and you pass it once the writer has said to, not to tidy up on your own initiative. Read the chapter first — read_book or get_chapter_fragments with includeText — because you are replacing what is there, not adding to it. Free — no credits.

save_console_preset

ChatGPT
Keep a tuning under a name so it can be recalled on any book. The usual way is fromBookId, which captures that book’s console exactly as it stands. Pass presetId to overwrite an existing preset instead of making another one. Free — no credits.

schedule_posts

ChatGPT
Write approved plans onto the content calendar, each carrying the plan it was approved with, so the post that publishes is the post that was approved. A plan that is not approved is refused by name rather than scheduled quietly. Nothing is generated or published now: each post is written and published at its own scheduled time, charged then. Free.

score_article_draft

ChatGPT
Grade a draft YOU wrote against the plan for its topic, before publishing it. Returns the score against the pass bar, each check with its weight, what to fix, and the draft with any link to a URL outside the plan's allowed list unwrapped (the sentence keeps its words and loses the link) — strippedLinks names every URL that went. Publishes nothing and stores nothing: fix what it names, score again, then pass the returned html to publish_to_wordpress. Free — no credits.

search_crm

ChatGPT
Find contacts, companies and deals by a piece of their text: a name, an email, a job title, a domain, an industry, a deal name or source. This is the tool for turning a name someone said into an id the other tools take. It matches substrings anywhere in those fields, case-insensitively, and returns a short row per hit rather than the whole record — follow with get_crm_record. An empty query returns nothing rather than the whole account. Free — no credits.

search_images

ChatGPT
Search the image library by prompt/filename/keywords. Prefer scope=mine. No permanent URLs returned.

search_studio_voice_library

ChatGPT
The provider’s entire public catalogue, not just the voices this account already holds — this is what the writer would otherwise leave the app to browse. Each result has a previewUrl, the provider’s own hosted mp3, so you can hand the writer a few to listen to. Free — a catalogue read and hosted audio, so auditioning a hundred voices spends nothing. A result is not castable. Each one carries libraryVoiceId and deliberately NO voiceId: the two kinds of id look identical, and casting a library id fails at generation time with "voice does not exist", long after a character is cast and a scene is written. Run add_studio_library_voice first and cast with the id IT returns. The response’s filters key lists the vocabulary this provider understands — read your filter values from there, not from memory. Only category is a closed set at the vendor; gender, age, accent and use case are free text and keep growing, and a value the catalogue has never heard of returns an EMPTY PAGE rather than an error. So an empty result means nothing matched, not that something is broken — widen the filters rather than reporting a fault. Two fields worth passing on to the writer: freeUsersAllowed: false means the voice will refuse an account on a free plan, and a non-null rate means its owner charges their own per-character rate on top of the model cost.

set_down_screenplay

ChatGPT
Takes the screenplay the author has ALREADY written in a book and records which part of it is a cue, a parenthetical, an action or a transition — the one thing standing between a finished script and a castable scene. It recognises; it does not rewrite, and it never invents a line. A fragment that already has lines is skipped, so running it twice changes nothing and a line corrected by hand is never overwritten. Use this when the text reads as a screenplay; use convert_prose_to_lines when it reads as narrative prose. Calls no model, so free — no credits. Needs the Acting Studio switched on.

set_funnel_sequence

ChatGPT
Replace the whole sequence — read it with get_funnel_sequence first or you will drop steps. It saves switched OFF, whatever you pass, because these messages go to real strangers who gave an address to a form, days apart, unattended. Preview it, then the owner enables it. Each step needs key (lower_snake_case), anchor ("signup" or "event"), offsetMinutes (negative is before, and an event anchor needs eventStartsAt on the sequence), subject and body. {{first_name}}, {{name}}, {{email}}, {{funnel_name}}, {{event_name}}, {{event_url}}, {{replay_url}} and the funnel’s own field keys are substituted; an unsubscribe link is added at queue time. Never invent a URL — leave a placeholder and say so. At most 12 steps. A step may also carry a condition, checked at send time rather than when it was scheduled, so a step can be written for people who have NOT done something yet: requireEvents (all must have happened), excludeEvents (none may have), stageIn / stageNotIn (the contact’s funnel stage). A step whose condition no longer holds is skipped, not delayed. Use it so a nudge about a step somebody has already taken is never sent. Free — no credits.

set_mini_blog_schedule

ChatGPT
Turn standing Mini Lesson generation on or off for a brand, and say how many a week and at what hour (UTC). Each lesson it writes costs 900 mana / 90 arcane when it runs, so tell the user what the week adds up to before you set it. The lessons are drafts, as they are everywhere else.

set_mini_lesson_image

ChatGPT
Set the picture at the top of a lesson. mode "reuse" (the default) takes the best relevant image the account already has and is free, reporting back rather than inventing one when nothing fits. "auto" reuses if it can and otherwise generates. "generate" always makes a new one. Those two cost 800 mana / 80 arcane per image. "url" uses an image the user gives you and is free. Say which you are about to do before you spend.

submit_brand_wizard

ChatGPT
Persist a confirmed brand/voice profile to ArcanoLabs, creating a brand the platform writes as. Only call this after the user has seen a sample and confirmed it sounds like them. The first brand on an account is free; each additional brand costs 100 mana. Set validateOnly to check what is missing without saving. A profile naming a brand that already exists returns that brand rather than duplicating it, unless force is set.

test_automation

ChatGPT
Answer "would this fire on a lead like this, and which condition is the one that fails?" against facts you supply. It writes nothing, fires nothing and charges nothing. Use it before enable_automation: the alternative is arming a rule and waiting for a real lead to discover a condition was written against a field that is never set. Free — no credits.

trash_wordpress_post

ChatGPT
Move a post on the site to its trash, where a person can restore it. Free — no credits. Use this to clear a superseded draft. ONE ACCOUNT CAN HAVE SEVERAL SITES: pass brand or brandId to say which site. A POST ID MEANS A DIFFERENT ARTICLE ON EVERY SITE, so pass expectedTitle: the post is read first and the removal refused if the title does not match. ONLY set force when the user has asked for the post to be destroyed outright — that empties it from the trash immediately and WordPress cannot undo it.

tune_console

ChatGPT
Turn the console knobs for one book: the class racks, the style and writing-type racks, a rack per character, location and event, the song rack, and the modulation matrix. This is the tuning a writer would otherwise do slider by slider, so when prose is coming out flat, too dense, too fast or with the wrong character carrying the scene, reach for this before rewriting anything — it changes every generation that follows instead of one passage. Send only what changes: everything you leave out stays as it is, and setting the dialogue rack never touches the scene rack. Numeric controls are 0-100; anything outside that is clamped. A control the console does not recognise is dropped rather than failing the save, and every drop comes back in a dropped list saying what was wrong with it, so read that rather than assuming a 200 means every knob took. Free — no credits.

unpublish_funnel

ChatGPT
Stop serving the public page. The funnel and everything it has already captured are kept, and publish_funnel puts it back at the same URL. Anyone following a link to it from an ad or an old email will get nothing, so say so before using this on a funnel that is in circulation. Free — no credits.

update_automation

ChatGPT
Replace a rule’s trigger, conditions or actions. The lists you send REPLACE the stored ones, so read it with get_automation first and send the whole list back with your change in it. A rule that is already armed stays armed, so an edit takes effect on the next matching event. Free — no credits.

update_crm_record

ChatGPT
Change the fields you send and leave the rest alone, on any kind of record. Same field names as create_crm_record. Read the record first if you are changing something a person may have edited since — this writes what you send, it does not merge two people's edits. To move a deal between stages use move_crm_deal, which keeps the pipeline and the deal in step. Free — no credits.

update_deck

ChatGPT
Replace a deck whole. Slides not sent are gone, so send the whole presentation, not a patch — to add one slide to a deck that is already up, use append_slide. Free — no credits.

update_fragment

ChatGPT
Change a single fragment: rewrite its text, change which template it is written under, set its template params, mute it out of the AI’s context (enabled: false leaves it visible but greyed out and excluded from generation), or override the console controls for this fragment alone. Replacing text keeps the old version in the fragment’s history, so an edit from a chat is as undoable as one made in the studio. Call get_console_controls first if you are setting controlOverrides. A param or control this fragment cannot take is dropped rather than failing the write, and comes back in droppedParams or droppedControls. Free — no credits.

update_image_metadata

ChatGPT
Update generation/category metadata for an owned library image.

update_mini_lesson_plan

ChatGPT
Save changes to the plan. This always returns it to draft, so an approved plan is always the one the user last reviewed. Free — no credits.

update_post_plan

ChatGPT
Change a plan and re-score it, or approve it so it can be scheduled. The link set may be REORDERED or SHORTENED and nothing else: a URL that is not already one of the brand's links is refused, because a post may only link to URLs the brand supplied. Add a new one in the Brand Wizard first. Free.

update_scene

ChatGPT
One tool for the whole life of a scene. Omit sceneId to create one; pass a sceneId to edit it. Pass delta (-1 up, 1 down) or position (1-based) to move it — it keeps its id and only its number changes, so a note that points at it still points at it. Pass remove: true to delete the container; its fragments stay and become unassigned, because merging two scenes should not take the pages with it. fragmentIds puts fragments in the scene, and a combined run always goes whole. characterNotes are keyed by character id and are kept only for characters the scene lists as present. Free — no credits.

update_story_bible

ChatGPT
Set the book-level fields: genre, mood, premise, plot, themes, style (pov/tense/pacing/tone). Only the fields you pass change; the rest are left alone, and no character or location is touched. Use upsert_story_entity for those. Free — no credits.

upsert_story_entity

ChatGPT
Add or change ONE character, location, object, event, method, motivation or plot point. Matched by id when you pass one, otherwise by name — so repeating a name updates that entity rather than duplicating it. Only the fields you pass change. Pass the fields that belong to the kind: character takes name/role/traits/relationships/notes, location and object take name/description/notes, event takes name/description/timestamp/order, method takes name/description/steps/limitations, motivation takes subject/motivation/reason/conflict, plotPoint takes description/section/order. Anything else is ignored. Free — no credits.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/arcanolabs"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/arcanolabs)