I spent years being the person other people called when their VICIdial broke. Somewhere in that time I noticed the same conversation happening over and over, and it was never really about the dialer.
It went like this. A team had outgrown running their own server. They wanted someone else to hold the pager. They looked at managed VICIdial providers, signed with one, and eighteen months later called me because they wanted to leave and could not.
The part nobody tells you at signup
VICIdial is open source. That is most of the reason it won. You can read it, patch it, and take it somewhere else.
Most managed providers fork it anyway. They add their own agent screen, their own admin, their own patches — sensible individually, and the accumulation is the problem. Your build stops being VICIdial and starts being theirs. The knowledge your team has stops transferring. And the exit, which was supposed to be a copy of a database and a config, becomes a migration project you have to justify to a board.
By the time anyone notices, the cost of leaving has quietly become the product. Not the uptime, not the support — the difficulty of getting out.
I have done enough of those extractions to find it genuinely annoying. The client is not paying me to make anything better. They are paying me to undo a decision that was reasonable when they made it.
So: stock, and root
VICIfast runs the unmodified open-source release. Not a fork, not a soft fork, not "mostly stock with a few improvements."
You get root over SSH. Not a support ticket that eventually reaches someone with root — root, on your server.
Both decisions are deliberate, and both cost me something. I cannot lock anyone in. I cannot ship a proprietary agent screen that becomes the reason you stay. If VICIfast stops being worth paying for, you copy your data and go, and everything your team knows still applies wherever you land.
That is the point. A hosting business that only survives because leaving is painful is not a business I want to run, and — more practically — it produces bad engineering. When your customers can leave easily, you have to be good at the actual job.
What I added instead
Everything I built sits alongside the codebase rather than inside it.
AI transcription and call scoring. Automated quality auditing, so QA reviews every call rather than a two percent sample. Text-to-speech for IVR prompts, because getting audio re-recorded is a week of nobody's life well spent. Two-factor auth on agent and admin logins. Whitelabeling. A firewall portal so you can add an IP yourself at 2am instead of opening a ticket.
Stock VICIdial has none of that, and it should not — it is a dialer, not an AI platform. But the gap is real, and filling it from outside means you still have VICIdial underneath.
Forty seconds
The provisioning time gets attention, so worth being plain about it: it is a consequence, not a feature.
If every build is stock, provisioning is deterministic. There is no per-customer configuration drift to reconcile, no bespoke patch set to reapply. The same automation runs the same way every time, which makes it fast — and, more importantly, makes it repeatable. The reason it takes forty seconds is the same reason your build looks like everyone else's, which is the same reason you can leave.
And the pricing
Flat per server. Not per agent.
Per-agent pricing punishes you for growing and, worse, makes your vendor's revenue go up exactly when your margins are under pressure. Every seasonal outbound operation I have worked with has at some point staffed a campaign more conservatively than the business case justified, because the licensing made the marginal agent expensive.
Charging for the server means scaling from five seats to five hundred does not change what you pay me. Bring your own carrier too. I would rather compete on running the platform well.
VICIfast is at vicifast.com — pricing is public and there is a seven-day trial if you would rather look than read.