Synadia

Inline protocol engine for realtime inspection and dynamic rulesets. The WAF for @nats_io

@nats_io Visit synadia dot com / protect to start a trial 🔓️ synadia.com/protect

@nats_io Same protocol, same semantics, just configurable policy in the middle. And it works with any OSS NATS deployment. See it action! (at our favorite burger chain: At-Least-Once-n-Out 🍔 ) youtu.be/qEiaM88UnVI

NATS auth checks who, but not what. In some enterprises - especially highly distributed architectures - this meant @nats_io deployments need exceptions or special clearance from security, risk, and compliance teams. No more!

Nice to see @synadia's insights get out in the world. (Datastar btw) synadia.com/insights

For years, edge systems summarized and aggregated because bandwidth gave you no choice. That constraint is gone. The habit isn't. Most teams still call "is the node up?" observability — and skip the layer that actually matters: end-to-end event traceability. The ability to answer *what happened to this specific event?* Not the average. Not the rollup. The actual event. Without it, your system can look perfectly healthy while quietly making decisions on signals that lost their meaning the moment they crossed the boundary. New post in the NATS Edge Eventing Architecture series — on why approximate observability is structurally inflating your MTTR, and what tracing-as-a-first-class-property looks like instead. 👇 synadia.com/blog/nats-edge… #EdgeComputing #Observability #NATS #EventDrivenArchitecture

Check out the video for what is possible, and as always feedback welcome. We want to build this with the ecosystem.

@nats_io One protocol, every agent youtu.be/TpybQ73E6jQ?si…

@nats_io The case for a heterogeneous agentic era 👉 synadia.com/blog/synadia-a…

The Synadia Agents SDK is here! The agentic era won't be one model or one framework — it'll be heterogeneous. That's why we've released an open-source SDK for unifying agents on one protocol.

Seven workshops. Three days. Live, with the experts. Following RethinkConn 2026 are seven workshops where you can in real-time ask "but what about my specific situation?" — and get an answer from a Synadia expert. → NATS Fundamentals → Hands on with Synadia Platform → Securing Your NATS Deployment → NATS at the Edge → Best Practices for Building Apps on JetStream → Operational Excellence with Insights → Building AI Agents on NATS The workshops happen live, only for registered attendees. Register: synadia.com/lp/rethinkconn… #NATS #JetStream #DistributedSystems

@nats_io Just answers that let you operate NATS with confidence. Start your trial today synadia.com/insights

Agent-ready... Monitoring and observability Triage and diagnosis Troubleshooting With Synadia Insights, you - and you AI agents - can see exactly what's happening inside your @nats_io system.

If your broker cluster has a "primary," you're already paying a tax you can't see on the invoice. Active-passive clustering is the comfortable default. It's familiar, it's well-documented, and most messaging broker docs lead with it. The trouble is that at edge scale, "comfortable default" quietly turns into your single largest cost driver — and most teams realize it too late to change cheaply. A few things I wanted to put on the table in Part 6 of the NATS Edge Eventing Architecture series: 👾 Passive nodes aren't idle. They're consuming compute, memory, and bandwidth to stay in sync. You're paying for N nodes and getting one node's worth of productive work. 🧗 Active-passive scales vertically. Every 10x in throughput becomes exponentially more expensive than the last, until you're forced into the architectural rewrite you should have done three quarters ago. 🤹 The load balancer tier you added "for HA" is its own failure domain — a critical-path component sized to peak single-node load. 😩 Failover isn't free. The latency spike between primary failure and passive promotion compounds downstream, especially when consumers have their own backlogs and time-sensitive work. Full-mesh clustering inverts the economics: every node is active, throughput scales linearly with nodes, and there's no primary-to-passive promotion sequence to wait through. Raft consensus handles the consistency story for durable streams, so you compose horizontal scale and correctness instead of trading them off. If you've ever sized a primary for "peak plus headroom" and watched the headroom evaporate, this one's for you. Full post: synadia.com/blog/nats-edge… #NATS #EdgeComputing #DistributedSystems #EventDrivenArchitecture #HighAvailability

📣 New NATS .NET releases v2.8.0-preview.3 and v3.0.0-preview.5 are out: 🌟 NATS Server v2.14 support, see release notes 🛠️ Dispose improvement, drain on dispose (opt-in) Please test this if you can github.com/nats-io/nats.n…

Cool.. github.com/open-telemetry…