What is AnyCable

AnyCable is a realtime server with delivery guarantees for Rails, Laravel, Node.js, Python, and any backend that can speak HTTP. It handles the hard parts of production WebSockets so your application does not have to: recovering missed messages after a connection blip, keeping users connected through deploys, tracking presence, and preserving message order.

It has been in production since 2017 and powers realtime features across healthtech, fintech, field services and IoT, collaboration tools, and SaaS products.

In a hurry? Jump to the Quick Start and pick your stack.

Why a separate realtime server

Most frameworks let you add WebSockets inside your application process. That is fine until production reality arrives:

• A user's network blinks for two seconds and the messages sent during that window are gone for good.
• You deploy, the process restarts, and every connected user is dropped at once.
• Traffic grows and a single application process becomes the bottleneck for every open connection.

AnyCable runs as a dedicated process next to your app. Your application stays the source of truth for authentication and business logic; AnyCable owns the connections. Because it is a separate process, a deploy of your app does not touch live connections, and because it is written in Go, a single instance holds over 820,000 idle connections at roughly 18 KB of RAM each (Pro), all while keeping latency low.

We measured these properties against popular Node.js setups (Socket.IO, uWebSockets.js). The full methodology and raw numbers are public; see the Node.js WebSocket comparison. The highlights, on a single 32 GB box:

• Capacity: 822,037 idle connections (Pro), where single-threaded Socket.IO topped out near 120,000.
• Memory: ~18 KB per connection (Pro), ~34 KB (OSS).
• Latency: 3 ms median, 11 ms p99 at 10,000 subscribers (Pro).
• Reliability: 100% of messages delivered under repeated network drops, where at-most-once setups lost 13-15%.
• Deploys: zero connections lost across application deploys that otherwise disconnected every user.

What you get

Capability	What it means
Delivery guarantees & recovery	Clients catch up on messages missed during a disconnect
Message ordering	Messages arrive in the order they were published
Presence	Track who is online in a stream, with join/leave events
Zero-downtime deploys	Connections survive application restarts
Efficiency at scale	Over 800,000 connections per instance in Go
Binary formats	Msgpack and Protobuf encoding to cut bandwidth (Pro)

See the Capabilities overview for how each one works and where it is documented.

How it fits your application

AnyCable supports two integration styles. You can mix them in one application.

1. Pub/sub (standalone). AnyCable handles connections and streams on its own. Your app authorizes clients with JWT and publishes messages over HTTP or to signed streams. No persistent connection between AnyCable and your backend is required. This is the quickest path for Node.js, Python, Laravel, and any HTTP backend.

2. RPC-backed. AnyCable delegates connection and subscription logic to your application over RPC. This is the default for Rails, where AnyCable is a drop-in replacement for Action Cable and reuses your existing channels.

Editions

AnyCable comes in three editions that share the same protocol and configuration, so you can move between them as your needs change.

• Open source. The full-featured Go server, free and self-hosted. MIT licensed.
• Pro. The same server with a denser memory model (further RAM reduction at scale), multi-node stream history, adaptive RPC concurrency, binary formats, and additional protocols (Apollo GraphQL, long polling, OCPP). It is a drop-in replacement for the open-source binary.
• AnyCable+. A managed service, so you do not run the server yourself. Free tier available.

Next steps

• Quick Start: get a server running and a client connected, by stack.
• Capabilities: the features that make AnyCable a realtime framework, not just a transport.
• Using with Rails: the Action Cable drop-in path.