# FlowFuse 2.31: Agentic Development Now in Open Beta, FlowFuse Expert Builds Your Industrial App on Edge Devices

> FlowFuse 2.31 puts FlowFuse Expert in open beta, building your industrial app for you on Cloud, Self-Hosted, and edge devices.

FlowFuse 2.31 lets FlowFuse Expert build your industrial application for you, now in open beta. Describe what you need, and Expert assembles it on your workspace.

## Let FlowFuse Expert build your industrial application

*FlowFuse Expert is our integrated AI assistant, in the website, the platform, and the immersive Node-RED editor.*

Describe what you want to build, an OEE dashboard, an MES handover screen, a UNS topic mapping, and FlowFuse Expert assembles the flow on your workspace for you. With 2.31 this moves out of soft launch into an open beta: switched on for your team on FlowFuse Cloud, no request needed. On Self-Hosted Enterprise, [contact us](/contact-us/) to get it set up; self-hosted setup is not automatic, it requires a provisioning token and configuring the bridge.

It now works on Remote Instances as well as Hosted Instances, so you get the same build-it-for-me experience across both. Agentic flow building requires the nr-assistant plugin at v0.16.0 or newer, and FlowFuse Expert will let you know when an update is available in your instance's immersive editor.

## Certified Nodes for industrial connectivity and AI

FlowFuse now offers certified nodes: vetted, FlowFuse-supported nodes you can add to your instances. At launch you can get:

- **RTSP**: bring live video streams into your flows, like camera feeds for monitoring, inspection, or feeding frames into a vision or AI pipeline.
- **OPC UA**: connect to your industrial equipment, the standard for talking to PLCs, SCADA systems, and machines on the plant floor.
- **AI nodes**: call leading AI models straight from your flows, with more providers following soon:
  - Anthropic
  - OpenAI
  - Gemini
  - Ollama, for models you run yourself<br />

These first LLM nodes handle single-shot, text-in/text-out calls; see the [LLM nodes documentation](/node-red/flowfuse/ai/llm-nodes/) for details. Multimodal input, conversation history, and tool calling are planned for follow-up iterations. They join the FlowFuse AI Nodes package we [shipped in 2.23](/changelog/2025/10/onnx-nodes/), which already includes the ONNX Inference, Image Classification, Object Detection, and Image Depth Estimation nodes for running vision and custom models on-device, with no external API calls.

Certified Nodes are available to FlowFuse Cloud customers as of this release, with Self-Hosted Enterprise support arriving in the next release (2.32). [Get in touch with sales](/contact-us/?subject=Certified%20Nodes) or your customer success manager, and we'll enable the ones your team needs.

## What else is new?

- **Control over where AI runs:** FlowFuse Cloud enables AI by default and you can toggle it per team. For Self-Hosted Enterprise, it is opt-in configurable for the platform via the `ai.enabled` configuration option. Once enabled, the same per team control is possible.
- **Custom CA certificates** are now documented for environments that need them.
- **Smaller fixes**: secure (`wss://`) connections to your MQTT broker now work from the editor, and custom packages with uppercase names no longer break pipeline deploys.

## A quick note from our team off-site in Greece

We spent part of this cycle together in Greece for a full-company off-site, and it shaped what you see (and do not yet see) in 2.31.

![The FlowFuse team in Greece](/blog/2026/06/images/offsite-team.jpg)

Getting everyone in one place let us reset our roadmaps and how we work, the kind of introspection that is hard to do day-to-day and fully remote. It is also why 2.31 is lighter on big features: a lot of this month went into foundational groundwork for what is coming next. We will share more soon.

---

For detailed breakdowns of each feature with additional visuals, visit our [changelog](/changelog/). For the complete list of everything included in FlowFuse 2.31, check out the [release notes](https://github.com/FlowFuse/flowfuse/releases).

If something in this release improves your workflow, or if there is still friction we can remove, please [share feedback or report issues regarding this release](mailto:contact@flowfuse.com?subject=Feedback%20on%202.31) with us.