Introduction

Welcome to Arcflux — the visual builder for agentic workflows.

Arcflux turns events like webhooks, schedules, chat messages, and live device streams into governed, multi-step AI automations. Instead of wiring together scripts and glue code, you build workflows visually: connect a trigger, add AI model steps that reason over your data, call the tools and services you already use, and gate sensitive actions behind human approval.

What you can build

A few things teams build with Arcflux:

• Automated triage and response — an inbound webhook or message starts a workflow that classifies the request with an AI model, looks up context, and takes action, such as creating a ticket, posting to Slack, or updating a record.
• Real-time monitoring and control — data from cameras, sensors, or machines streams in over MQTT, NATS, or AMQP; a workflow analyzes it and reacts, escalating to a human when needed.
• Scheduled and on-demand assistants — recurring jobs or chat-triggered agents that gather information, reason over it, and produce results or take action across your stack.

Core concepts

A few ideas show up throughout Arcflux:

• Triggers start a workflow. A trigger can be a webhook, a schedule, a chat message, or a real-time device or message stream.
• Models are the AI steps that reason over your inputs. You choose the provider for each step — OpenAI, Anthropic, Google, xAI, or Groq — so you can pick the right model for the job.
• Tools and integrations are the actions a workflow can take. Arcflux connects to external services through native integrations and MCP servers, so your workflows can read from and write to the systems you already use.
• Approvals put a human in the loop. You can require sign-off before a workflow takes a sensitive action.
• Workflows tie it together — a sequence of steps, with branching and per-workflow scoping of data and credentials, plus run-level history so you can see exactly what happened.

Why Arcflux

• Build visually. Compose triggers, models, tools, and approvals without maintaining glue code.
• Use any model. Mix and match providers across steps instead of committing to one.
• Connect everything. Native integrations and MCP give your workflows a broad reach.
• Work in real time. Ingest live device and message streams alongside webhooks and schedules.
• Stay in control. Human approvals, per-workflow scoping, and run-level auditing keep automated actions governed.

How a workflow runs

At a high level, every workflow follows the same shape: a trigger fires, the workflow moves through its steps — model calls, tool calls, branches, and approval gates — and produces an outcome or takes action. Each run is recorded, so you can review the inputs, outputs, and actions taken.

Next steps

• Quickstart — build and run your first workflow.
• Core concepts — go deeper on triggers, models, tools, and approvals.
• Connect a model provider — add your OpenAI, Anthropic, Google, xAI, or Groq credentials.
• Integrations and MCP — connect the services your workflows act on.
• Real-time streams — ingest MQTT, NATS, and AMQP data.

Getting help

Need a hand? Reach us at info@arcflux.ai.