Skip to main content

Frequently Asked Questions

Basics​

What is RPA?​

Robotic Process Automation is software that operates other software the way a person does. Instead of needing an API, a robot opens the application, clicks the button, reads the screen and types the value — so it can automate systems that were never designed to be automated.

Robomotion adds a visual flow, so the sequence is something you can see and reason about, and AI nodes, so steps needing judgement rather than rules can be handled too.

How is this different from a script?​

A script is fine for one task on one machine. What you tend to need after that is the part around the script: scheduling, credentials that are not in the source, a run history, retries, someone other than the author being able to change it, and a place to look when it breaks at 3am. That surrounding machinery is what an RPA platform is.

What is the difference between a flow and a robot?​

A flow is the design — nodes and the connections between them. A robot is the worker that executes it. Flows do nothing until a robot picks them up. See Your first flow.

Do I need to be able to program?​

No for most work — flows are built by dragging nodes and setting properties. Some help, though: the Function node runs JavaScript, and every template in the library uses it for the glue between nodes. If you can write a few lines of JavaScript you will move noticeably faster.

Which operating systems are supported?​

Windows, macOS, Linux and Raspberry Pi. See Install Robot.

Where do I start?​

The tutorial path — seven short tutorials, in order, each teaching one idea.

Robots and running flows​

Does my robot have to be connected?​

To run a flow, yes. A robot executes the work, so it must be online when a schedule fires or a trigger arrives. Robots can live on your computer or on a VPS — see Windows VPS or Linux VPS for unattended setups.

Cloud Run is the exception: Robomotion starts a robot in its own cluster on demand, runs the flow and shuts it down, billed by the minute.

What is the difference between development, production and application robots?​

A development robot runs flows only from the Flow Designer and cannot be scheduled — it is for building and testing. A production robot is the unattended one that schedules and triggers use. An application robot backs a published application. See Robot.

Does a robot need a public IP or an open firewall port?​

No. A robot makes an outbound TLS connection and keeps it open, so it needs no inbound access at all. A laptop behind NAT works exactly like a datacentre host. Outbound 443 to your workspace host and the package registry is enough.

My flow works on my desktop but fails on the server. Why?​

Nine times in ten: the flow drives a GUI and the server has no interactive desktop session. Headless browser work, API calls, database and file operations need no display; headed browser automation and any GUI automation do. On Linux use X11, VNC or Xvfb; on Windows the robot must run in a logged-on session, because a service in session 0 has no desktop.

Can one robot run several flows at once?​

A robot runs one flow at a time. For concurrency, add robots — that is what the robot count in your plan or licence governs.

How do I stop a running flow?​

From Admin Console → Jobs, or POST /v1/jobs.stop in the API.

Credentials and security​

Is it secure to store my credentials?​

Credentials go in a Vault and are encrypted in your browser, with 256-bit AES, before they are uploaded. What reaches the service is ciphertext, under a key derived from your password and your vault secret key — neither of which the service has ever seen.

Can Robomotion see my passwords?​

No. Sign-in uses SRP-4096, so your password never leaves your device in any form; the server stores a verifier it cannot reverse. Vault items are encrypted before upload. See Security Model.

What happens if I forget my password?​

You can reset it and carry on using the workspace, but your vault contents are gone — the key that decrypted them was derived from the password, and nobody at Robomotion can decrypt them for you, by design.

The protection is to share any vault that matters with a second person, at the time you create it. A vault two people can open survives one of them losing a password; sharing it afterwards is impossible.

Do my business documents pass through Robomotion?​

Not unless a flow explicitly sends them. Flows execute on your robot, so an invoice it reads and posts to your ERP never reaches the control plane. What the control plane learns is that a job ran, on which robot, for how long, and whether it succeeded.

Check three exceptions: queue item payloads, uploaded artifacts, and anything your flow writes to its own logs.

Can I run everything inside my own network?​

Yes — On-Prem runs the whole control plane on your infrastructure. If the requirement is about data location rather than who operates the software, Private Cloud usually satisfies it with far less ongoing effort. See Deployment Topologies.

Is there an audit log?​

Yes — every consequential action with the acting user, a timestamp and before/after state. See Audit, or /v1/audit.logs for export.

Building flows​

Are custom nodes supported?​

Yes. The package system supports Go, Python, Java and C#, and roboctl can scaffold a package from an OpenAPI specification. For logic inside a flow, the Function node runs JavaScript without any packaging.

How do I pass data between nodes?​

Through the message object — one JSON object that travels the flow, which each node can read from and add to. It is the concept everything else rests on: Passing data between nodes.

How do I loop over a list?​

A ForEach node with a Label and a GoTo — see Repeating work.

How do I handle errors?​

Three ways, and picking the right one is most of the skill: Continue On Error per node for optional steps, a Catch node to react and retry, or letting it fail when a human should look. See Handling failure.

My browser selector keeps breaking. What should I do?​

Prefer, in order: a stable id, a data-* attribute, visible text, then structure. A selector on a generated class like css-1x7fj3k will break at the site's next deploy, and that is the single most common cause of a flow that "used to work". Use Inspect to get a starting point, then make it robust.

Can I reuse part of a flow?​

Yes — Subflows are callable, versioned units.

Can flows be version controlled?​

Flows have built-in versioning with diffing, and Git integration is available for teams that want flows in their own repository.

AI​

What can AI Agents do that a flow cannot?​

Handle steps with no decidable rule — classifying an invoice, deciding whether a support message is a complaint, reading an unfamiliar document layout. See AI Agents, Skills and MCP.

What is the difference between a tool and a skill?​

A tool is a capability — a node an agent may call itself. A skill is judgement — written instruction about how your organisation does something. An agent with a Sheets tool can write to a spreadsheet; an agent with your month-end skill knows which spreadsheet and what to do about a row that does not balance.

Which models can I use?​

Claude, OpenAI, Gemini, Ollama for local models, and OpenRouter for many behind one interface.

Can I run models locally so prompts stay in my network?​

Yes — use Ollama. Same reasoning as deployment topology, applied to prompts.

What if the agent gets it wrong?​

Design for it. Route on confidence and send the uncertain cases to a human queue rather than acting on them. A confidence threshold plus a queue turns an unreliable step into a reliable process.

API and integration​

Is there an API?​

Yes — the API Playground lists every endpoint a token can reach, with runnable requests and code in nine languages. The machine-readable description is at

How do I authenticate against the API?​

A bearer token from Admin Console → API. See Authentication.

What is the difference between a read-only and a read-write token?​

Read-only tokens reach listing and reading endpoints; read-write tokens also create, update and delete. The full grid is in Permissions.

Can I trigger a flow from outside Robomotion?​

Several ways: POST /v1/flows.run in the API, an HTTP trigger, a webhook, or a Form for a human-supplied start.

Is there a CLI?​

Two, for different jobs. robomotion-deskbot is the robot — it connects and runs flows. roboctl manages the workspace from a terminal: robots, flows, jobs, schedules, users, licences, audit, and package building. See Command-Line Interface.

Troubleshooting​

A property is empty at runtime. Why?​

Almost always because the node that sets it runs later than the node reading it, or does not run at all on the path taken. Add a Debug node before the failing node and look at the message — see Passing data between nodes.

Where do I see why a flow failed?​

Admin Console → Jobs keeps the error and logs for every failed run. In the Designer, use Debugging.

My Switch node did nothing and the flow just stopped.​

A Switch whose conditions are all false emits on no port at all, so execution ends silently. Add a final true catch-all branch, or make the last condition the general case.

Inspect is not working.​

Install the Chrome extension. It is needed for building flows only, never for the robot runtime.

I cannot find a node for the service I need.​

Check Packages first — there are over 170. If it is genuinely missing, build one: the package system supports Go, Python, Java and C#, and roboctl can scaffold from an OpenAPI specification.

Still stuck?​