Your AI Operator.
Monitors. Explains. Acts.
Smartunit Operator follows live work across events, tools, memory, and approvals. Ask what changed, see why, then approve what happens next.
From Flow to Agent in
Three Steps
Start where your work already happens. Add the agent nodes, connect a model and tools, then run the flow and inspect every step in Studio.
Install the Nodes
Add the agent package to Node-RED. You get the building blocks for agents, inputs, skills, models, memory, connections, behaviour, and budgets.
$ npm install node-red-contrib-agent
Connect Model and Tools
Choose the AI model, write the agent instructions, and expose your existing flows as tools the agent can call when it needs data or action.
Run and Inspect
Send a message, trigger the flow, or schedule a run. Studio shows the session timeline, tool calls, memory updates, alerts, usage, cost, and final response.
"Explain this event, check the latest data, and summarize what changed."
Built for
Live Operations
Most agent builders stop at chat. Smartunit Operator puts agents inside Node-RED, where they can use your flows, remember context, call tools, and show every run in Studio.
Connect to Your Work
Use the Node-RED flows you already have. Let the agent reach APIs, databases, files, webhooks, notifications, industrial protocols, and internal tools through the same environment your team knows.
Run Close to the Data
Keep the runtime in your Node-RED and Smartunit environment. Point it at the AI provider you trust, from a hosted model to a local or controlled endpoint.
Ask With Context
Ask what happened, what changed, or what to check next. The agent can use conversation history, saved facts, tool results, and live flow data to produce a useful answer.
Act Through Tools
Turn existing flows into agent tools. The agent can call named actions with parameters, then Node-RED handles the actual work: lookups, diagnostics, messages, tickets, or downstream automation.
Monitor Runs
Trigger agents from schedules, events, buttons, or incoming messages. Studio shows sessions, alerts, token use, cost, and activity so you can see what is running and what needs attention.
Inspect Every Step
Open a session and follow the timeline: user request, agent thinking, tool calls, tool responses, memory reads and writes, usage, warnings, and final response.
Free and Custom Agent Flows
Start freely, or tell us the workflow and we will build an agent flow that fits it.
Community
Start building useful agents inside Node-RED for free.
- Build agents directly in Node-RED
- Connect the AI model you prefer
- Let agents call your existing flows as tools
- Remember useful facts between sessions
- See what happened during each run
- Set token or spend limits
Custom
Tell us the workflow and we will build the agent flow for you.
- Turn your workflow into an agent flow
- Connect your tools, data, and systems
- Choose the right model and guardrails
- Set up dashboards for runs, memory, and alerts
- Track usage, cost, and agent activity
- Tune it with your team after launch
Frequently Asked Questions
Common questions from builders, automation teams, operations leaders, and IT managers.
Smartunit Operator is a Node-RED agent runtime with AI Agent Studio on top. The runtime lets a flow call an LLM, use Node-RED skills as tools, keep memory, emit events, and track usage. Studio gives your team a dashboard for agents, sessions, events, memory, alerts, and costs.
The free node gives you the agent runtime: agent, input, skill, model, connection, behaviour, memory, and budget nodes. You can build multi-step agents, connect OpenAI-compatible models, expose Node-RED flows as callable tools, store facts and events, choose context modes, and set token or spend caps.
Studio makes the runtime visible. It shows a multi-agent dashboard, agent flow canvas, session and event explorer, memory browser, alerts, token usage, and cost totals. It is the place to inspect what the agent did, which tools it called, what it remembered, and where errors or warnings happened.
The runtime connects to OpenAI-compatible APIs. That can be OpenAI, OpenRouter, compatible hosted providers, or a local endpoint such as Ollama if it exposes the right API shape. You configure the base URL, API key, primary model, optional summary model, and temperature.
Tools are built as Node-RED skills. A skill defines actions, descriptions, parameters, and outputs, then the agent can call those actions during a run. Because the tool lives in Node-RED, it can reach whatever your flow can reach: APIs, databases, files, webhooks, industrial protocols, notifications, or internal systems.
The agent can store facts and append event logs in Smartunit datastores. Memory can be scoped to a session or a user, and history can run as full conversation, rolling summary, last N messages, or stateless mode. Studio lets you inspect memory records and the event timeline for each session.
The runtime runs inside your Node-RED and Smartunit environment, but model requests go to whichever LLM endpoint you configure. For private deployments you can point it at a local or controlled provider; for cloud models, the data sent depends on your prompts, tools, memory settings, and provider policy.
We are not offering packaged paid plans right now. You can use the free Node-RED agent runtime without artificial agent or source limits. For custom work, tell us the workflow, tools, data sources, and model setup you need, and we will build the agent flow and Studio setup around it.
Yes. We can start with one focused workflow, such as incident explanation, event triage, shift summaries, or human-approved actions, then tune the agent flow around the result.