Moving off paper fax means replacing three things at once: the machine, the analog line behind it, and the filing habits built around both. A fax server handles the first two by sending and receiving over your network instead of over copper. The third one is where most migrations stall, because a fax that arrives as a PDF has to land somewhere a person will actually look.
This is a practical guide to that move. We build ICTFax, a FreeSWITCH-based fax server, and the failure modes below are ones we’ve watched customers hit rather than ones we’ve read about.
Why the analog line is the part forcing your hand
Most teams that still fax do it because a hospital, a court, or an insurer on the other end still does. That’s a fine reason to keep faxing. It’s not a reason to keep the copper.
Carriers in the United States have been retiring analog lines for years, and the replacement offers are almost always digital. When your fax machine ends up behind a VoIP line, it usually stops working properly, and the failure looks random: some pages go through, long documents fail at page four, and confirmations come back mangled. We covered the timeline and what to do about it in the piece on what happens when your analog fax line is retired.
The reason is worth understanding, because it explains why a fax server is the fix rather than a workaround.
Fax over VoIP breaks for a specific, fixable reason
Fax was designed for a circuit that stays open and delivers every tone in order. VoIP does neither. Audio is compressed, packets arrive late or not at all, and the fax protocol treats a missing tone as a transmission error, retrains, and eventually gives up.
Two things make it work again:
T.38 stops sending the fax as audio. Instead it decodes the fax data at your end, sends it as data packets, and re-creates the tones at the far end. Lost packets can be re-sent, so ordinary network jitter stops mattering. This is what you want, and it’s what a fax server speaks natively.
G.711 passthrough sends the fax as uncompressed audio and hopes the network behaves. It works on a clean local network and falls apart across the public internet. Treat it as a fallback, not a plan.
If your carrier does not support T.38, that’s a carrier problem, and it’s worth solving before anything else. Changing SIP providers is cheaper than debugging fax failures for a year.
What actually changes when a fax server replaces the machine
The obvious change is that documents arrive as PDFs by email or in a web inbox. The changes that matter more are the ones nobody demos.
Faxes stop sitting in a tray. A page that prints in a shared corridor is readable by anyone walking past. A fax routed to a named inbox is not. For anyone handling health or legal records this single change closes the most common real-world exposure.
You get a delivery record. Every transmission carries a status, a timestamp, page count, and the remote station identifier. When someone claims they never received the referral, you can answer with a record instead of a memory.
Sending becomes something software can do. Once faxing is an API call, your existing systems can send without a human. That’s the change that removes the most labour, and it’s the one most teams postpone.
Retention becomes a decision. Paper faxes get shredded on an informal schedule. Digital ones sit on a disk until someone sets a policy. That’s an improvement only if you actually set one.
The compliance piece, stated plainly
If you handle protected health information in the United States, faxing is permitted and always has been. HIPAA does not ban fax. What it asks is that you control access, keep an audit trail, and have an agreement in place with any vendor that touches the data.
That last clause is where hosted fax services and self-hosted servers differ. A cloud fax provider processes your documents, so you need a business associate agreement with them and you inherit whatever their retention and breach history looks like. A fax server you run yourself never hands the documents to a third party at all, which removes the agreement question rather than answering it. Our HIPAA fax explainer covers the requirements in detail.
The failure that actually causes breaches is duller than encryption: a fax sent to the wrong number. We wrote about that in why most fax breaches are a wrong number and most fines are a missing risk analysis. A fax server helps here in a way paper cannot, because the destination comes from a directory rather than from someone’s handwriting.
Open source or commercial: pick on operations, not price
ICTFax exists in two forms, and the honest way to choose between them has nothing to do with features.
The community edition is free and open source under GPL-3.0. Source is on GitHub, there’s an official Docker image, and the documentation lives at ictfax.org. It’s the whole application, not a trial. If you have someone comfortable with Linux, SIP, and a container runtime, this is a genuinely complete option and costs nothing.
The commercial edition on this site adds white label branding, multi-tenant billing so you can resell fax as a service, and a support contract. That last item is what people are usually buying. If a fax failing at 2am is somebody’s emergency, having a number to call is worth the licence.
Both are built on the same core. ICTCore handles the call and document handling, with FreeSWITCH underneath for the telephony and T.38 negotiation. There’s no crippled build; the difference is packaging and support.
Our comparison of ICTFax and HylaFAX is the useful read if you’re weighing open source options against each other, since HylaFAX is the other serious answer in that space.
A migration that works
The order below matters more than the speed. Teams that try to switch everything in a weekend end up running both systems for six months anyway.
- Confirm T.38 with your SIP carrier before you buy anything. Ask them directly, and ask for a test number. This single question prevents most failed migrations.
- Stand the server up and test with your own numbers. Send to a machine you control, both directions, including a long document. Ten pages will expose problems that one page hides.
- Port the fax number last. Run the new server on a new number while the old line still works. You want the fallback available when the first real fax fails.
- Decide where inbound faxes land before you go live. A shared inbox that nobody owns is worse than the paper tray it replaced. Route by DID to a named person or a department queue.
- Write the retention rule down. How long documents stay, who can read them, and what deletes them. Do this while the system is small.
- Connect the sending side to whatever system generates the documents. Email to fax is the quick win, and the email to fax setup guide walks through it. The API is the bigger one.
Expect the number port to be the slow step. Carriers take days to weeks, and that timeline is outside your control, which is exactly why it belongs after everything else works.
What this costs compared with what you’re paying
An analog line plus a machine plus toner and paper is rarely a line item anyone reviews, which is why it survives. Add up the line rental, per-page consumables, the maintenance contract, and the staff time spent walking to the machine, and the annual figure surprises most people.
A fax server replaces that with a server you already have or a modest cloud instance, plus per-minute SIP charges that are a fraction of an analog line’s rental. For the community edition, the software cost is zero. The real cost is the hours of whoever sets it up, which is why the commercial edition exists.
Where self-hosting clearly wins is volume. Cloud fax services charge per page, so a busy department’s bill scales with usage forever. A server’s cost barely moves between 500 and 50,000 pages a month. If you send more than a few thousand pages a month, run the comparison; the crossover usually arrives sooner than expected. Our ICTFax against eFax comparison lays out that maths.
FAQ
Can I keep my existing fax number? Yes. Numbers port to a SIP carrier the same way voice numbers do. Keep the old line active until the port completes and the new setup is tested.
Do I still need a fax machine? No. Users send from email, a web interface, or an API call, and inbound faxes arrive as PDFs. Some teams keep one machine for a year out of caution, and almost all of them unplug it eventually.
Is fax over IP reliable enough for medical records? With T.38 and a carrier that supports it properly, yes, and it’s more reliable than a machine that jams. Without T.38 it isn’t, which is why the carrier question comes first.
What if the far end still uses an old fax machine? That’s the normal case and it works fine. Your server negotiates with their machine over the phone network exactly as another machine would.
Is the open source version limited? No. The community edition is the full application under GPL-3.0. The commercial edition adds white label branding, tenant billing, and support, not core fax capability.
How long does a migration take? A single-department move is typically two to four weeks, and most of that is waiting on the number port. The technical setup is a day or two once T.38 is confirmed.
Where do the documents live? On your server, under your control. Set a retention policy early rather than discovering three years of PDFs later.
If you want to see the interface before deciding anything, the live demo is open, and the community edition is one Docker command away if you’d rather run it yourself first.