Vaultbeat

Roadmap

What's coming, what we're weighing up, and what we've decided never to build. Already shipped? That's the changelog.

No dates on this page, on purpose. Vaultbeat is built by one person alongside a university degree, so throughput swings a lot across the year. A date posted here would be a guess dressed up as a promise. Order is roughly priority; everything above Considering is committed.

Being built now

In progress

Committed work, actively in progress.

Live Activity for your partner’s sleep

Their sleep state on your Lock Screen and Dynamic Island, pushed live — no opening the app to check whether they made it to bed.

Instant sync after your agent writes

Today, when your AI logs a meal or a lift, your iPhone picks it up the next time you open Vaultbeat. This makes it land within seconds via a silent push.

Under consideration — for your agent

Considering

Ideas we think are right but have not committed to. Feedback moves things up this list faster than anything else.

Webhooks — stop making your agent poll

Push new health data to your agent as it arrives, so it can act on a bad night without being asked first.

Aggregate and trend queries

One MCP call that answers "how has my resting heart rate moved over two years" instead of returning thousands of raw points for your agent to chew through.

Joint queries across both partners

Every other health MCP server is single-person. Vaultbeat already holds two people's data with explicit consent, so couple-level analysis is something only it can do.

Docker image and documented HTTP fallback

For people running their agent on a home server or NAS rather than a laptop.

Under consideration — in the app

Considering

Same status: wanted, not promised.

A heads-up to your partner before PMS

The most requested thing in cycle-tracking communities is some version of "I want him to know without me having to bring it up". The prediction and the sharing pipeline both already exist.

Symptom-versus-cycle correlation views

The data is already recorded and already syncing; what is missing is the chart that makes a pattern visible.

More Apple Health types

Blood oxygen, noise exposure, stand hours — each one is also a new thing your agent can reason about.

Apple Watch app and an iPad reader

Log from your wrist; read on a bigger screen.

Last-synced indicator

The top complaint about couples apps generally is silent sync failure. You should be able to see at a glance that their phone is actually still talking to yours.

Deliberately not building

By design

These are not backlog items we never got to. They are decisions, and they are what makes the rest of the product possible.

Cloud AI or a built-in chatbot

The whole premise is that intelligence runs on a machine you own. The moment we run a model over your health data in our cloud, we have to be able to read it — and then end-to-end encryption is marketing rather than architecture.

A social layer — friends, feeds, leaderboards

Two people is the boundary. Health data gets less safe with every additional viewer.

A knowledge base inside the app

Your agent already has somewhere to keep notes, and it is better at it than we would be. Vaultbeat stays a data source.

Selling data, ads, or training on your data

We only ever hold ciphertext, so this is not a policy we could quietly reverse later — it is not technically available to us.

Android — for now

One person maintains this. Two platforms at this size means two mediocre apps. Honest answer: not soon.

Want something on this list?

Requests from people actually using the app reorder this page faster than anything else. Open an issue or email — a human engineer reads every one.