Your First Chapter
Connect Alcove to Claude, ChatGPT or Cursor, then write and lock your first chapter without leaving the assistant you already use.
This takes you from nothing to your first locked chapter, entirely inside the AI assistant you already write in. There is no Alcove app to open.
The system proposes; the writer steers.
1. Connect Alcove to your assistant
Alcove is a remote MCP server:
https://mcp.alcove.fi/mcpClaude Desktop / claude.ai — Settings → Connectors → Add custom connector, paste the URL, sign in when the browser opens.
Claude Code —
claude mcp add --transport http alcove https://mcp.alcove.fi/mcpThen /mcp, pick Alcove, authenticate.
ChatGPT, Cursor, and others — add the same URL wherever the client accepts a remote MCP server and complete the OAuth sign-in.
There is no API key to paste anywhere. Full detail: MCP & your AI.
2. Create or Remix a cartridge
Ask your assistant to make one. A Cartridge is a reusable story world — cast, tone, rules, configuration — so you never begin from a blank prompt.
"Create a new Alcove cartridge for a rain-soaked harbour city where the tide is a political actor."
You can also Remix an existing cartridge into a private attributed copy, or import a draft you already have and let Alcove propose starting canon around it. See Import & export.
3. Open a Save File
A Save File is your timeline in that world — chapters, locked canon, cast, codex and route all live in it. Ask your assistant to create one and name it. See Save Files.
4. Ask what happens next
Rather than starting cold, ask Alcove for next moves — concrete options derived from your open threads and locked canon. Pick one, steer it, or state your own intent. See the core loop.
5. Write the chapter
Write it in the assistant. It can read your canon first, so chapter 40 knows what happened in chapter 3 — that is the whole point of connecting.
Then have it submit the chapter as a proposal. Submitting is free and changes nothing yet.
You can also write somewhere else entirely — another model, or by hand — and submit the finished text. Alcove does not care who wrote it.
6. Read what the checks found
Alcove runs continuity, intended-change, story-rule, safety and configured quality checks against that exact version, and returns what they found. Baseline continuity and safety are on every plan.
This is the part worth slowing down for. The checks are the product — the writing is the commodity.
7. Review the proposed canon changes
Before anything is permanent, Alcove shows the state changes the chapter implies: who died, what was revealed, which promise closed. Keep, revise, or drop each one.
Generated drafts never change your world on their own. They produce proposals; your approval is the gate.
8. Canon Lock
When you are happy, Canon Lock it. Your assistant will ask before doing this — it is never automatic.
Draft is not canon until you lock it.
Everything before this point is safe to rewrite or abandon. See Canon Lock.
9. Continue
Your cast, codex, timeline and routes catch up from the new canon, and the next moves are recalculated. The loop begins again.
Your story remembers itself.
Where to go next
- The full cycle: the core loop
- Every tool your assistant gets: MCP & your AI
- Building blocks: Cartridges and Save Files
- Costs: Plans & billing
What is Alcove?
A narrative operating system for long-form fiction — write chapters in your own AI assistant, have them checked against your canon, and decide what becomes true.
The core loop
Decide the next change, bring in a chapter, run checks, review its canon changes, and Canon Lock only what you approve.