Why a UiPath proxy decides whether the job finishes or faults
An RPA workflow that opens a browser is just another network client, and the target site scores its IP long before the first selector resolves. Datacenter ranges sit at a fraud score of 75–100 out of 100, which is why a robot that passed testing suddenly reports "element not found" — the page it actually loaded was a captcha. IPs from mobile carrier networks score 0–15 on the same scale, the highest trust level of any proxy type. In practice that means the queue drains overnight instead of dropping fifty faulted transactions into the morning report.
Retries are where the money leaks. Each faulted transaction re-runs the whole invoke, doubles robot runtime and, on cloud licensing, burns budget on requests that were never going to be answered. Route the same traffic through a carrier IP and the runtime you pay for produces information instead of error screenshots.
UiPath Studio proxy configuration: build on the IP you will ship
Most teams apply the uipath studio proxy at the system level — Windows internet settings or the HTTP_PROXY and HTTPS_PROXY variables the process inherits — and point package feeds at the same route through the uipath.settings file. Login:password authentication works from any machine, which matters when a developer moves between the office and a home connection; IP whitelisting is the cleaner option on a workstation with a fixed address. Both methods are supported, so a change of location never means rewriting the configuration or opening a ticket.
Record selectors through the same operator and city the unattended robot will use later. Mobile pages, currencies and consent banners differ by carrier and region, and a workflow debugged on one route often breaks on another. Matching them from day one removes the classic "works in Studio, fails in production" gap.
UiPath Robot proxy for unattended and long-running jobs
A uipath robot proxy has a harder task than the development one: hold a stable identity for hours with nobody watching. On the Regular plan the device is rented whole, so a sticky session keeps a single IP for the entire login-to-logout sequence, and the IP changes only on command — by link or by timer. Sessions stop dying mid-transaction, and the robot no longer has to re-authenticate ten times per shift.
Pro tip: drop the rotation link into an HTTP Request activity at the end of your Get Transaction Item loop. Every queue item then starts from a fresh carrier IP without a single change to the business logic, and rate limiting stops being a reason to split one job into five batches.
| Workflow type | Session mode | What it protects |
|---|
| Portal login, form submission, invoice download | Sticky session | The session token survives to the end of the transaction |
| Bulk collection across many pages | Rotation by link, per batch | Request limits never trigger on a single address |
| Overnight monitoring, price or SERP checks | Rotation by timer | The job stays alive without an operator on shift |
| Low-risk lookups, internal tools | Automatic rotation every 2–5 minutes (Lite) | Lowest cost per port for simple flows |
One port per robot, one profile per account
When robots drive accounts rather than public pages, the rule is rigid: one port, one browser profile, one account. Sharing a single route between several accounts on the same platform links them together, and a fleet of ten accounts can be lost to one flagged login. Dedicated ports keep the failure contained — one suspended account no longer drags the rest of the farm with it.
SOCKS5 support is what makes this work with antidetect browsers such as Multilogin, AdsPower, Dolphin Anty or Octo, which the robot then controls through the standard ip:port:login:password string. That combination gives the network identity and the browser identity separately, which is the industry standard for multi-account automation.
What a mobile IP will not fix
The carrier IP solves the network layer and nothing above it. Canvas, WebGL and font fingerprints still need an antidetect profile; robotic timing — zero-delay clicks, perfectly linear navigation — still looks like a script, so randomized delays inside the workflow remain necessary. Language, timezone and currency must match the country of the port, otherwise the whole setup contradicts itself.
Speed has limits worth planning around: expect 50–300 ms latency, 5–50 Mbit/s, and roughly 1–5 parallel threads per device. That is invisible for form filling, portal work and data collection, but heavy file transfers or video belong on your normal route. Sizing the fleet by threads instead of by hope keeps the schedule realistic.
Lite or Regular: matching the port to the fleet
| Plan | Access | IP control | Fits |
|---|
| Lite | Shared device, up to 5 users | Automatic rotation every 2–5 minutes, not adjustable | Public data, monitoring, low-risk requests |
| Regular | Dedicated device for the whole rental | Full functionality: sticky, rotation by link, rotation by timer, device reboot | Account work, long unattended jobs, Orchestrator fleets |
Billing is per port for 1, 7 or 30 days, with 24 hours as the minimum period and no per-GB metering on either plan — unlike residential offers priced by traffic, a heavy scraping month does not change the invoice. Prices depend on the country and operator you pick and are shown on the plan page. Cost per robot becomes a line item you can forecast a quarter ahead.
A free server proxy is available through the widget on the site and is handy for verifying that your configuration and authentication work at all; mobile ports themselves are paid, with no trial. Cashback lands as promo credits on the internal balance after a rental ends, and refunds follow the refund and replacement policy — full within the first hour after access is issued, later minus the time used, with a replacement offered first when the problem is technical. Support runs around the clock with a four-hour target for the first reply, so a stalled overnight queue gets attention before the business day starts.
Budgeting a UiPath proxy is therefore an infrastructure decision, not a gamble: fixed price per port, predictable thread count, and a rotation model chosen per workflow. That is what turns a UiPath proxy from an extra expense into the reason your automation reports success instead of exceptions.