Closed beta

What your team knows, your AI should know too.

One knowledge base for the whole house — series bibles, style rules, the prompts your best editor wrote — readable by every writer and by the AI sitting next to them. Some of it readable by everyone. Some of it not.

Why we built it

The house style lives in three people's heads.

Most writing teams we talked to run on a documents folder, a wiki, and whatever chat assistant each writer happens to prefer. It works until it does not, and it fails in three specific ways.

🧩

The AI has read none of it

Every writer re-explains the world, the voice and the rules at the start of every session. The answers differ because the briefings differ.

🌀

Nobody knows which version is real

A wiki page, a pinned message and a Google Doc all describe the same character, and all three are out of date in different directions.

🚪

Everything walks out with anyone

The prompt library that took two years to write is one folder download away from being on somebody else's laptop, and there is no record it ever happened.

The workspace

Three sections, not ten tabs.

The team area used to be ten flat tabs. It is now grouped by what you are doing there — getting work done, feeding the team's shared brain, or running the place — and each pane has a URL you can send someone.

Work
Overview
The review queue across every book, plus what was approved recently and which chapters have not moved in a week.
Books
Team books through their whole life — active, finished, and a trash that holds for sixty days. A personal book can be moved in.
Team Brain
Knowledge Base
The shared bible as a file tree, with semantic search over it. Import Word and PDF; edit a single file without leaving the tab.
Skills / Prompts
The reusable instructions your team has written, enabled or disabled per team, each one showing whether it is confidential.
AI Memory
What the AI has learned about your world, held in a review queue. A manager confirms, edits or deletes before it becomes shared truth.
Rules
Team-level rules for the AI Coach — mandatory or advisory, and scoped to prose, screenplays or both.
Admin
Members & Seats
Who is in, what they can do, and how many seats are used against how many you hold.
Usage & Billing
Where the team's AI usage went, and the seat summary. Visible to the owner and admins only.
Audit
Every read of a team skill, who read it, from where, and whether it was withheld or throttled.
Settings
Rename the team, hand ownership to someone else, or close it down.
Confidential skills

Usable by everyone. Readable by you.

Mark a skill confidential and it keeps working — a writer's AI conversation still uses it, still follows it, still produces the house voice. What changes is that the text itself never reaches their device, and only owners and admins can open it.

  • No copy syncs down: the member's machine holds the name and the description, nothing more
  • A marker is embedded in the text, and any AI reply that reproduces it raises an alert
  • A daily read budget, plus a burst rule for anyone opening eight different skills in five minutes
Skill access · this week
Northlight Press
Series voice guide · used in chat 42×
! Bulk read throttled · one member today
Full text read · two admins
Confidentiality acknowledged 11 of 12

What this stops and what it does not: it stops casual copying, a leaving employee packing the folder, and bulk export. It does not stop someone deliberately coaxing the content out of a model, an owner or admin choosing to share it, or a photograph of a screen. We would rather you know that before you buy than after. The classification, the differentiated permissions and the contemporaneous access log are also the three things courts tend to look for when deciding whether information was actually kept secret — though what that is worth in your jurisdiction is a question for your own counsel, not for us.

Roles and seats

Writing and approving are different jobs.

Four roles, and they stack — one person can be an admin and a writer at once. A reviewer reads, comments and approves without occupying a seat.

Role Write Use team AI Manage members Takes a seat
OwnerOne per team. The only role that can transfer ownership or close the team.
AdminInvites, removes, manages skills and rules. Cannot remove the owner.
WriterCreates and writes team books. Sees their own books plus the knowledge base.
ReviewerOpens team books in a read-only workspace to comment and approve.

During the closed beta we set your seats up with you rather than asking you to buy them from a page. Tell us the shape of the team on the form below and we will size it.

Chapter review

A state on every chapter, and a queue for the editor.

Not a project management tool bolted on the side. Three states on the chapter itself, and one table that shows the whole book.

✍️

Drafting

A target word count and a state the writer sets. The editor sidebar and the manuscript table both show where each chapter stands.

🔍

In review

Submitted chapters collect in the team overview across every book, so an editor works one queue instead of opening eight projects.

Approved, or back to drafting

Approve it or send it back. Either way it is a state on the chapter, visible to the person who wrote it without anyone having to send a message.

Closed beta

We are onboarding teams one at a time.

There is no self-serve checkout for this yet, and that is deliberate. We set each team up by hand, watch how it goes, and fix what breaks before the next one. Tell us about yours and we will get back to you.

We use this to contact you about the beta and nothing else. No newsletter unless you ask for one, and you can have the record deleted by writing to [email protected].

Questions

What people ask first.

Why is it a closed beta? +
Because a team plan is not a feature, it is somebody's working week. We would rather set up ten houses carefully and learn what actually breaks than open a checkout and find out from support tickets.
What does it cost? +
The team plan itself is on the pricing page. Seats for the people you invite are sized with you during the beta rather than bought from a page — tell us the shape of the team and we will quote it before anything is charged.
Can a reviewer see everything? +
A reviewer opens team books in a read-only workspace with comments, and reads the knowledge base. They do not consume the team's AI allowance and they do not take a seat. Confidential skills stay closed to them like they do to any non-manager.
What happens to our books if we stop paying? +
Nothing is deleted. The workspace goes read-only, the books stay yours, and everything exports in the formats it always did — Word, EPUB, PDF, Markdown, or the whole project as an archive.
Can our writers use their own AI tools? +
Yes, over Slima MCP. A confidential skill stays confidential across that path too: the assistant gets the name and the description, not the text.
Do you train models on our manuscripts? +
No. Work in Slima is never used to train models and is never shown to other users. The narrow cases where anyone here could see a manuscript are written out in the privacy policy.

Tell us about your team.

Two minutes on the form. A reply from a person, not a sequence.