Arcflux is now in alpha

Today we're opening Arcflux to its first users. Alpha is live, and the waitlist is the front door.

Arcflux is the visual builder for governed, device-driven AI workflows. You model a workflow on a canvas, trigger it from anywhere — including live device streams — and govern every run before it spends a credit. No orchestration code.

Here's what that means in practice.

What you can build today

Workflows as a graph. Open Studio, drag nodes onto a canvas, connect them. Triggers, LLM agents, tool calls, branches, knowledge lookups, and human approvals are all nodes. The graph is the program — there's nothing to deploy.

Triggers from anywhere. Start a workflow from a webhook, a schedule, a chat message, the API, or — the part most platforms don't have — a live device stream over MQTT, NATS, or AMQP. Point a workflow at a broker topic the way you'd add a webhook. No bridge service to subscribe, normalize, and babysit. One workflow can listen to several sources at once.

Governance before spend. Every event passes a policy gate before a run is created: dedupe, cooldown, throttle, condition, and always-on safety checks. A skipped event costs zero credits, so a noisy sensor or an alert storm never burns your quota. Tune each policy in shadow mode against real traffic, then flip it to enforce.

Your models, your tools. Switch the model per step — Claude, GPT, Gemini, Grok, and Groq — without rewriting the flow. Bring any MCP server or native integration (Linear, Slack, Stripe, databases, and more); tools are scoped per workflow.

Humans in the loop. Workflows pause at approval gates. The agent does the legwork; a person approves the risky step. Every run is auditable, step by step.

What "alpha" means

We'd rather be clear than oversell, so:

• It's early. Expect rough edges and fast iteration. Your feedback shapes what ships next.
• It's waitlist-gated. We're onboarding in batches to keep the experience tight. Join the waitlist and we'll bring you in.
• Basic is free. Start with 100 credits a month — enough to build and run your first workflows. A credit is roughly a cent of LLM usage; a simple run costs about one. Paid plans go live at launch; waitlist members get access first.
• The roadmap is public in spirit. CoAP and MAVLink are coming for constrained devices and drones. SSO/SAML and dedicated data isolation are on the enterprise track.

Who it's for

If your events come from the physical world — sensors, vehicles, machines, robots — and you've been putting off "connect an AI to this" because of the plumbing, alpha is for you. It's also for anyone who wants governed, multi-step AI on top of the webhooks and schedules they already run, without writing an orchestration layer to do it.

Get in

Join the waitlist. We're onboarding in batches and would love to see what you build.

This is the start. The interesting automation of the next few years starts at a sensor — and now there's a governed place to put it.