AI assistants
Your scheduler, inside your assistant.
Weekwright speaks the Model Context Protocol, the open standard behind Claude's connectors and ChatGPT's apps. One endpoint, every assistant that speaks it — reading your week and drafting changes through the same tool layer the dashboard runs on, under a much shorter leash.
- Category
- AI assistants
- Setup
- 1 minute, no key to copy
- Who sets it up
- Anyone, for their own access
- Plans
- Every plan, including the free tier.
What you get
Claude, ChatGPT and MCP clients + Weekwright.
- Ask the questions you would otherwise click through: who is short on Saturday, would this week pass our compliance rules, who could cover Tom's shift and why, how are hours spread across the team this month.
- Let it draft. With changes allowed it can create shifts, move shifts and post open shifts — and it cannot publish the week, so nothing reaches your team until you say so.
- Works wherever you already work: the Claude apps on desktop, web and mobile, Claude Code, ChatGPT, Cursor. It is one standard endpoint, not a per-vendor integration we have to build one at a time.
- One sign-in, no credential to manage. Approve the connection, pick the organization, done. There is no key to paste, store or rotate — and every change writes an audit entry naming you.
Setup
How it works.
- 1
Add Weekwright as a connector in your assistant and give it the endpoint: https://weekwright.com/api/mcp. In the Claude apps that is Settings → Connectors → add a custom one; in ChatGPT and Cursor it is the equivalent MCP server field.
- 2
Sign in to Weekwright and approve. The screen asks two things: which organization, and whether the assistant may make changes. Changes are off by default.
- 3
Ask it something. It has 29 operations out of the product's ~137 — a hand-curated list, four of which write.
- 4
Disconnect whenever you like; the assistant's next call is refused immediately.
Honest limits
What this does not do.
Every integration has edges. Better you read them here than discover them the week you go live.
- It cannot publish a week, approve time off or a swap, invite anyone, change roles, touch billing, or delete anything. Those are outside the curated list entirely, so no setting enables them — an assistant drafts and a human decides.
- It never has more access than the person who connected it, and it tracks that person: a demotion narrows the assistant the same day, and leaving the organization stops it working without anyone remembering it existed.
- One organization per connection, chosen at approval. A request naming a different one is refused rather than redirected. To switch, disconnect and reconnect.
- Claude is the client we have verified end to end. ChatGPT, Cursor and the rest speak the same protocol and the endpoint is the right shape for them, but we will not claim a client we have not tested — this line gets updated as each one is. We are also not listed in any assistant's directory yet, so you add the connector by URL rather than finding us in a gallery.
Common questions
Claude, ChatGPT and MCP clients FAQ.
Which assistants does this work with?
Anything that speaks MCP, which is the point of using a standard rather than building one integration per vendor: the Claude apps on desktop, web and mobile, Claude Code, ChatGPT, Cursor, and the growing list of clients that have adopted it. Claude is the one we have verified end to end so far. You add Weekwright as a custom connector by URL in whichever you use.
What is MCP?
The Model Context Protocol: an open standard that lets an AI assistant call tools in an outside system. Claude's connectors are built on it, as are ChatGPT's apps. It means one endpoint serves every assistant that speaks it, rather than a separate integration per vendor — and when a new assistant adopts it, Weekwright works there on day one with nothing to build.
Can the assistant change the schedule without me knowing?
It can only change anything if you ticked 'allow changes' when you connected it, and even then it cannot publish — so nothing reaches your team's phones without a human publishing it. Every change writes an audit entry naming you and broadcasts in realtime, so a manager with the schedule open sees it appear.
Is my team's data being used to train a model?
Weekwright does not send your data anywhere for training. What the assistant does with what it reads is governed by the assistant vendor's own terms — Anthropic's for Claude — which is worth reading before you connect any tool to any system.
Does an employee get one too?
Yes, bounded to what an employee can already do: their own shifts, their own availability, requesting time off. They cannot see pay, other people's notes, or anything an employee cannot see in the app.
Is there a separate API if I want to build something myself?
There is no public developer API, deliberately. Where a machine credential is genuinely needed, a scoped integration key is bound to one organization with an explicit default-deny tool allowlist and a role ceiling — and it works against this same endpoint, for clients that cannot do a sign-in flow.
One tool for the schedule, wired into the rest of your stack.
Free for a single location, no credit card. Every plan, including the free tier.