Common NATS Financial Services Questions
NATS has become a popular piece of architecture for financial services events and streaming. We see NATS being downloaded and deployed in a plethora of financial services categories from FinTechs to Banking, Trading Desks, Payment Processing… NATS is everywhere. I recently created an education module to help developers and architects in financial services better understand why and how to use NATS for a variety of common use cases.
Below you’ll find the answers to questions financial services professionals asked on their journey to getting started and future-proofing their architectures with NATS.
Can the NATS Server run on Android as a leaf node?
In terms of capacity requirements, yes. NATS has a tiny footprint, less than 30 megabytes binary. You can run it anywhere you want to (It runs very well on Raspberry PIs, and many other devices with lesser capacity than an Android phone). Now, we have not done tests running NATS on an Android device, but there is no reason why it shouldn’t be able to run in theory. The only real concern would be the networking capabilities/limitations of the device, and the Android OS itself.
.
The other challenge is that in order to install it on an Android device, you would have to find some proprietary (manual) mechanism, as NATS is not an application that would be available on the Google AppStore.
Is there NATS JetStream support for stream processing APIs?
Yes, there's a few things you can do. We have an out-of-the-box product called Wombat (https://wombat.dev), which is basically a fork of Benthos. You can use Wombat to ingest any kind of events on it and do your transformations, mapping, and any kind of processing logic you want to do. Wombat is open sourced.
We also have the Connector Framework that offers connectivity to various data sources as well well as various processing logic capabilities. The Connector Framework is currently available on the Synadia Cloud.
Beyond this you can have various other event stream processing products like Apache Flink be able to leverage NATS streaming capabilities and apply processing logic to the event stream. You could use Flink to look at time or volume based windows, apply all manner of logic and look for patterns in the event stream.
The bottomline is you can use multiple tools to achieve stream analysis with NATS and JetStream.
What's the difference between Synadia Cloud and Core NATS?
Core NATS is Open Source. Synadia Cloud is the multi-tennant platform that wraps all manner of IP around NATS i.e. deploying NATS to Kubernetes, monitoring NATS, applying security constraints, offering workload management, connectivity etc. You can sign up on Synadia Cloud and use it, but you share the infrastructure with everybody else. Synadia Cloud is a great thing to start playing with to help get things figured out, and to get your knowledge base built up.
That being said, most financial institutions would opt for the more “private” Synadia offerings, as they usually do not want to share infrastructure with other companies. To that effect Synadia has the Platform offering (which is a private packaging of the Synadia Cloud) that can be deployed within the customer’s environment and either be managed by Synadia or the customer themselves.
Is there an AI connector for NATS? Can Agentic AI workflows submit inbound prompts to NATS?
Let’s start with the second question first. Agentic AI is on Synadia’s radar, and we are actively making steps towards an offering. In the interim we are partnering with companies like n8n where we offer connectors into the platform that customers can use to build AI Agents and both ingest and dispatch events on the NATS infrastructure. As an example, with n8n you can trigger the execution of an agent via an incoming event on a NATS subject, and can send the reply over another NATS subject. Over time we will look at expanding connector support to other platforms, so stay tuned for that.
With regards to “AI connectors” for NATS, we have discussed exploring MCP capabilities for NATS as well as ways to interact with various AI models. Additionally, Workload Management can be used to execute AI logic at the edge, and NATS itself can be used to coordinate training cycles for models at the edge.
Lastly, NATS can be used as part of broader Agentic AI Initiatives, by acting as the “nervous system” for the signaling required to distribute prompts to various agents and aggregating agent responses.
How do you define the difference between eventing and streaming from your perspective? Which part of NATS would you say accomplishes each part?
Eventing deals with the event routing and distribution. This traditionally falls in the category that products like TIBCO, Solace, IBM etc. traditionally do. Whereas eventing is more of a fire and forget model (although some platforms offer durable subscription to ensure no event loss), Streaming in general deals more with ensuring that events are delivered in chronological order, they can be replayed and can be retained/stored for historical querying. Kafka is a great example of streaming.
Core NATS offers eventing, and JetStream offers streaming. As such where you would need to use 2 or more products to cover all use event use cases, NATS gives you all capabilities on one platform !
How do I get started?
I would just urge you to simply download the NATS server, and start playing with it. It's easy to use. It's small. It has a good, comprehensive set of APIs out-of-the-box (NATS CLI). You don't have to code to get going with it. Then you go from there. If you want to use Docker and/or Kubernetes you can do that as well, because NATS is also packaged as a docker container, but you can also run it natively. You can then develop your knowledge about how the NATS platform works (event routing, subjects, streaming etc.).
Once you get the basic functionality, you can start building architectures that are more scalable, that are broader (i.e. look at clustering, leaf nodes etc), and provide different functionality for your needs. We also have a NATS API to build microservices that can be monitored and managed as part of the infrastructure itself. So that's one thing you can do to build solutions that are visible to the Control Plane i.e. here's my list of microservices running on NATS, and I can see them and manage them. That's the next step I would do when you want to start coding. However you don't have to start with code. Just start with a typical NATS CLI, and you can get a lot done with that.
Then you can explore the broader Synadia capabilities i.e. Control Plane, connectors, workloads etc.
