habits.chat
Get early access

MCP server

Track your habits from the assistant you already use

habits.chat exposes its whole tool registry over the Model Context Protocol. Connect Claude — or anything else that speaks MCP — and it can check you in, pull your streaks and answer questions about your history, under scoped access you grant and can revoke.

Get early access See how it works

Why it matters here

The habit gets logged where you already are

Habit tracking dies of friction. Not big friction — the fifteen seconds between remembering and having the right app open, which is exactly long enough to decide it doesn't matter tonight.

If you already have an assistant open — on your phone, in your editor, on your desktop at 11pm — that's the lowest-friction surface that will ever exist for you, and it isn't ours. Trying to make you switch to our window would lose to the sofa every time. So the tools go to where you are instead.

There's a second reason, which is that it's your data. An account whose contents can only be reached through one vendor's chat box isn't really yours. MCP is how we make that concrete rather than a claim in a privacy policy.

Claude, connected to habits.chat Chat

did I read last night? and what's my longest streak been

Called 2 tools

  • get_history(habit: "reading", window: "3d")
  • get_streak(habit: "reading", mode: "longest")

Yes — 25 minutes, logged at 22:40. Current streak is 11 nights; your longest was 34, ending in March when you were travelling.

A read-only token can answer this. It cannot create, edit or delete anything.

What you grant

Scoped, attributed, revocable

Connecting an assistant to months of personal habit data is a real decision. It should be a granular one.

Read-only by default

A new token can answer questions and nothing else. Write access — logging, editing, deleting — is granted separately and can be pulled back without touching the read scope.

Per-habit scoping

Grant an assistant your reading and exercise habits and not the one you're quitting. Sensitive habits stay out of any connected client unless you explicitly include them.

Expiring tokens

Set a token to expire in a week, a month, or never. Trying a client for a fortnight shouldn't mean a credential that outlives your interest in it.

Every call logged

Which client, which tool, which arguments, when. The MCP log is the same log the app writes to, so you can see exactly what an assistant did and undo it.

Revoke without losing anything

Pull a token and the check-ins it made stay. Disconnecting a client is not a reason to lose three months of history.

And a plain export

MCP is for assistants. For everything else there's a full export of every habit and check-in you've ever logged, in a format you can open.

The registry

The same tools the app uses

Not a reduced API surface bolted on afterwards. The MCP server exposes the registry the board and the built-in chat already call.

Exposed over MCP

create_habit set_target set_check_in log_check_in log_slip log_missed get_streak get_history find_pattern correlate_habits compare_periods summarize_progress set_nudge export_data

A tool added for a screen is available over MCP the same day. There is no second implementation to fall behind.

Questions

About connected assistants

Which assistants can connect?

Any MCP client. Claude is the one we test against first, and it's what most people on the waitlist say they already use, but nothing about the server is specific to it.

Does connecting an assistant send my habits to that vendor?

It sends whatever the assistant requests in order to answer you — the same as any tool call. That's why scoping is per-habit and read-only by default: you decide what's reachable before anything is asked.

Can I use the MCP server without using your app?

Yes. Some people will only ever touch habits.chat through an assistant, and that's a legitimate way to use it. The web app is where you'd go to see the board, manage tokens and export.

What happens if a connected assistant logs something wrong?

Every write is attributed to the client that made it and appears in your history as an edit, so you can find it and undo it. Nothing is silently overwritten.

Is this an extra charge?

No. The MCP server is part of the plan, not an add-on. Locking the API behind a higher tier would undercut the entire point of exposing it.

Early access

Get a token when it opens

Tell us which assistant you use and we'll make sure it's tested against before you get access.

  • MCP server included on every plan
  • Read-only tokens available from day one
  • Founding accounts lock their price for life

No card required. Unsubscribe any time.

Your habits, reachable from anywhere you work

Founding pricing is locked for accounts opened during early access.

Get early access See how it works

Founding accounts lock their price for life. No card required.