Ably for AI Applications

The realtime layer for Temporal-powered AI applications

Connect Temporal Workflows directly to clients with reliable, realtime transport. Ably carries live, long-running AI output to users at scale - so you don’t have to manage the infrastructure yourself.

Start building free Jump to docs

Durable execution is only half the battle

As AI scales, the connection becomes the challenge. Production applications don't just need to complete work reliably – they need to keep users connected to that work as it unfolds. This is where Ably comes in: a realtime transport layer, resilient and proven at billions of connections.

Keep sessions connected

Clients stay on the same realtime channel even as workers restart, scale, or hand off long-running work.

Stream without the extra hops

Deliver tokens and results directly from workers to clients, keeping latency low throughout long-running interactions.

Deliver once, even after a drop

When newer steps supersede older ones, the client only ever receives the one that matters — no replays, no duplicates to reconcile on reconnect.

Why durable execution turns into a chokepoint

Temporal's event log becomes responsible for two very different jobs: recording the work and delivering it.

Streaming degrades under real usage: The only path from worker to client is a multi-hop route back through the Temporal log, which was designed for discrete events - not for streaming tokens at volume. The cost: throttled output when you need it most.
Retries that corrupt what the user sees: When work is retried, the system has to reconcile new output with what may already have reached the client - risking duplicates, gaps, or broken streams.

Why durable execution turns into a chokepoint image

How Ably fixes the chokepoint

Ably drops in alongside Temporal to give workers a dedicated realtime channel to publish directly to clients. The result is:

Lower-latency streaming: Token output goes straight to the client, so response time isn't held hostage by execution overhead. Critical for conversational AI.
Retries that never reach the UI: AI Transport brackets output by workflow step. When Temporal retries a step, the new output supersedes the partial one on the channel, so however many retries happen, the user sees one clean response.
Delivery that scales independently: Client connections scale with your user base; execution scales with your workload. Neither one has to grow to keep pace with the other.

What's included, so your team doesn't have to manage it

Reduce engineering time spent on infrastructure. Ably works with Temporal to give you the continuity, responsiveness, and resilience that your AI product needs.

Capability	Temporal alone	Temporal + Ably AI Transport
Workflow durability	Temporal alone Yes Survives crashes and restarts	Temporal + Ably AI Transport Yes Survives crashes and restarts
Automatic retries and replay	Temporal alone Yes Activities retry automatically on failure	Temporal + Ably AI Transport Yes Activities retry automatically on failure
Delivering workflow output to a live client	Temporal alone No No built-in path to a frontend	Temporal + Ably AI Transport Yes Output is streamed as the workflow progresses
Session survives tab close / reconnect	Temporal alone No Client has no way to resume	Temporal + Ably AI Transport Yes Client resumes from last position on reconnect
Multi-device continuity	Temporal alone No Each device must poll independently	Temporal + Ably AI Transport Yes Any device joins the same session and catches up
User visibility into long-running workflow	Temporal alone No Users see nothing until the workflow completes	Temporal + Ably AI Transport Yes Agent state and progress stream to clients in real time
Human-in-the-loop signal (approve, redirect, cancel)	Temporal alone No Signals work backend-to-backend but don't reach the user	Temporal + Ably AI Transport Yes Signals travel both ways over the durable session
Multi-worker fan-in to one user session	Temporal alone No Each worker runs independently with no shared user session	Temporal + Ably AI Transport Yes Multiple workers publish to one session, user sees one stream
Enterprise proxy traversal for delivery	Temporal alone No Temporal has no transport layer for browser delivery	Temporal + Ably AI Transport Yes Automatically falls back to HTTP streaming if WebSocket is blocked
Session diagnostics for user-facing side	Temporal alone No Execution history only; nothing for the frontend session	Temporal + Ably AI Transport Yes Connection state, presence, and delivery events surfaced in real time
Replay-safe message delivery	Temporal alone No Workflow code outside activities fires again on every replay	Temporal + Ably AI Transport Yes Step IDs with AI Transport deduplicate delivery across retries and replays