Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Quick Start

Install and run Invar in 5 minutes

Quick Start

Invar speaks the Redis wire protocol, but it's not Redis: there's no in-memory store to configure and no cluster to stand up. One process, one writer, and your data lands in object storage (or disk, for local dev) instead of RAM. This page gets you from nothing to a running instance in about a minute, then to a production-shaped instance backed by S3.

If you haven't seen it yet, the Invar guide overview covers the single-writer model and why it works this way — worth a skim before you go further than "does it run."

Install

Docker (all platforms):

docker pull ghcr.io/hardpointlabs/invar:latest

For convenience, we also publish container images to Docker Hub, however downloads from many cloud and CI environments are aggressively rate limited

Homebrew (macOS, Apple Silicon only):

brew install --cask hardpointlabs/tap/invar

This installs the same binary that runs inside the container, signed and notarized. Once installed, invar is on your $PATH and takes the exact same flags as the container image; every command below works by dropping the docker run ... wrapper and calling invar directly. Intel Macs and Linux desktops aren't covered by the cask yet- use Docker in those environments.

Invar ships as a single executable with a minimal dependency footprint. The same binary runs locally against a disk, or in production against S3-compatible object storage.

Invar requires glibc to be present in the host environment. Invar also ships without a compiled root CA bundle. On MacOS it defers to the system keychain; on Linux, this is typically shipped as a separate package (e.g. ca-certificates on Debian/Ubuntu).

Run it locally

This starts Invar with the fjall backend, which persists to a local path instead of object storage. It's the fastest way to try Invar or run it in CI, and it needs no cloud credentials:

docker run -v /tmp/invar:/tmp/invar -p 6379:6379 -it \
  ghcr.io/hardpointlabs/invar:latest \
  --backend fjall --path /tmp/invar --redis

Or, if you installed via Homebrew:

--redis tells Invar to serve RESP on :6379. --path is where fjall keeps its data — with Docker that's bind-mounted to /tmp/invaron the host so it survives container restarts; with the native binary it just writes there directly.

With that running, connect using any Redis client. redis-cli works as-is:

That JSON.* support is the part that isn't "just Redis" — Invar speaks the same document commands you'd expect from RedisJSON, on top of the same wire protocol. There's more on this in Usage.

Run it against object storage

Swap the backend to slate (Invar's SlateDB-backed mode) and point it at a bucket. This is the same code path you'd run in production — see Operations for what that looks like at deployment scale:

Or natively:

Invar picks up the standard AWS_* environment variables for credentials — nothing Invar-specific needed there. --prefix(default /invar) namespaces the keys Invar writes within the bucket, useful if you're sharing a bucket across multiple instances or tenants.

What's next

  • Usage — what's different from talking to real Redis: JSON commands, transaction caveats, scripting, and the command groups that don't apply here.

  • Operations — running this for real: compute deployment (k8s, Fly.io, serverless) and configuring storage and caching.

  • CLI reference — every flag and environment variable.

  • Redis command support — the full compatibility matrix.

Last updated