Local-first sync engines

A Replicache alternative for one-shot moves rather than multiplayer

Optimistic local writes, an app that keeps working offline, and data that shows up on the user’s other device.

What Replicache is

In its own terms.

A client-side sync framework from Rocicorp. Your app writes to a local cache through mutators, Replicache pushes them to your backend and pulls updates back, and you implement push and pull on your server against your own database. Rocicorp now builds Zero, a query-driven successor with its own sync server.

Side by side

Where they part ways.

	s3nd	Replicache
What it is	A CLI, a library and React hooks over your own S3 bucket	Client sync framework with push and pull on your backend
Where the data lives	Your bucket: S3, R2, MinIO, Scaleway, Wasabi	Your backend database, cached on the client
Both ends online at once	No. The code is redeemed later, until it expires	No
From inside your web app	Yes, through one route on your server	Yes
Accounts	None. The code is the whole handshake	Whatever your backend requires
Hosting	Your bucket, plus your server when a browser takes part	Your backend; Zero adds a sync server
Price	Free. You pay your storage provider for a few objects that expire	Free
License	MIT	Source available; Zero is Apache-2.0

The differences

What actually changes.

A framework with a backend contract versus a primitive

Replicache asks you to implement push and pull, keep a version per client and per space, and store everything in a backend database. s3nd asks for a bucket and gives you put and get.

Multiplayer versus a personal transfer

Replicache is built for many clients editing shared data in realtime with rebase and conflict resolution. s3nd is built for one person carrying their own data to their next device.