Skip to main content
Stacklane

Real-time features, no third-party vendor required, on the database you already run.

Real-time is the substrate, not a feature flag. We've shipped operator dashboards with hundreds of live data sources, multi-tenant SaaS with presence, and AI agent UIs streaming tokens, most of them on the database you already run plus the WebSocket layer you already trust. When the workload genuinely needs a high-concurrency runtime, we reach for one.

What we build

  • Database-native notifications as the bus

    For the 80% case (operator dashboards, in-app notifications, presence in a multi-tenant workspace), the database itself can carry the realtime traffic. No separate pub/sub layer to run, no Pusher subscription, no Ably bill. The database is already there; it's already the source of truth; we make it the bus too.

  • WebSockets that share the API's data contracts

    WebSocket routes share the same data contracts as the HTTP API, so a field rename surfaces as a build error in every client, not as a 4am production bug. Authentication on the upgrade request, not after. Backpressure handled at the framework level. Reconnect logic baked into the client SDK so dropped connections don't surface to users.

  • High-concurrency runtime where it earns it

    Collaboration tools, presence at high concurrency, soft-realtime gaming, workloads where a lightweight process-per-connection model is the right fit. We reach for a high-concurrency runtime where the numbers say it ships faster and runs cheaper than the default; we leave it on the table where the default is enough.

  • Reconnect + replay on the client

    Lost connection, dropped message, missed window during a deploy, all handled by the client SDK. Messages have monotonic IDs; the client resumes from its last seen ID after reconnect. No 'just reload the page' UX.

  • Backpressure that doesn't crash the server

    Slow clients don't take down the realtime path for everyone else. Per-connection send buffers with bounded size; messages drop or coalesce when a client falls behind. The dashboard surfaces which clients are degraded so operators can intervene.

  • Observability built into the realtime path

    Per-channel message rates, per-client latency, drop reasons, all instrumented and visible. Realtime systems fail invisibly without this; we make the failures legible from day one.

Where this fits

  1. You need realtime in a SaaS dashboard and the current plan is 'pay Pusher €1,200/month and hope', which doesn't scale economically.

  2. You're shipping a collaborative product and the prototype works for 5 users but falls over at 50 because there's no presence layer.

  3. Your AI agent UI needs to stream tokens to the client and the current implementation is server-sent-events bolted to a REST endpoint with no reconnect.

Tech stack

  • Postgres LISTEN/NOTIFY
  • Elysia
  • WebSockets
  • Phoenix
  • Elixir

Want this for your team?

30 minutes to scope what you need. No pitch deck, no obligation. We tell you straight whether Stacklane fits.

Book a Free Call