How to choose a VICIdial host — an evaluation checklist

Every managed VICIdial host runs the same open-source dialer. The software is not the differentiator. What separates one host from another is operational: how fast you get a box, how much of it you actually control, what happens when something breaks, and what it costs you to leave.

This is a checklist you can take to any vendor, including us. Each section starts with the short answer, then explains why the question is worth asking. If you would rather see the checklist already applied to named vendors, we keep head-to-head comparisons with other managed VICIdial hosts.

What does a managed VICIdial host actually take off your plate?

It takes over the install and the recurring Linux and Asterisk maintenance. It does not take over your dialing.

The work a managed host absorbs is the same list every VICIdial operator eventually writes down:

• The initial build — Asterisk, Apache, MariaDB, the VICIdial schema, the screen-managed daemons, and the dozen settings that have to agree with each other
• Asterisk and VICIdial version upgrades
• OS and package patching
• TLS certificate issue and renewal, and the debugging when renewal fails
• Firewall rules, re-cut every time an agent moves or a carrier changes a signalling IP
• Backups you have actually tested a restore from
• The 3am incident when the keepalive stops and nobody notices until the morning shift logs in

None of that is hard once. All of it is recurring. Managed hosting is a decision to pay a flat monthly fee instead of staffing that work — which is a good trade or a bad one depending on whether you already employ someone who does it well. There is a section at the end on when self-hosting is still the right call.

How fast can you get a server, and what does that tell you?

Provisioning time is the cleanest proxy for how automated a host really is. Ask for a number, then test it yourself.

Broadly there are three shapes:

• Manual builds. You email or open a ticket, a technician builds the box, you get credentials hours later or the next business day.
• Scripted builds. An installer runs unattended, but a human still kicks it off and checks it. Typically half an hour to an hour.
• Fully automated builds. Payment triggers the build; nobody touches it. Under a minute is achievable.

The number matters less than what it implies. A host that can build a box in under a minute can also rebuild yours after a failure, hand you a second box for a two-week campaign, and give you a demo server while a prospect is still on the phone. A host that needs a technician for every build needs a technician for every recovery too, and that technician has working hours.

Test it, do not take it on faith — including from us. VICIfast's managed VICIdial hosting builds a server in about 40 seconds, and the free 7-day trial makes that a test you can run before you pay anything.

Who holds root on the box?

You should. If a host will not give you root SSH, you are renting a black box rather than a server.

Root matters the first time you need something the dashboard does not expose: a custom dial-plan context, an AGI script, reading /var/log/asterisk when a call path misbehaves, or pulling your own data export on your own schedule.

Ask:

• Can I add my own SSH public key without opening a ticket?
• Can I log in as root, or only as a restricted user?
• If I break something with a root change, what does support do — fix it, refuse, or restore from backup?

A reasonable host says yes to the first two and "we will restore you from backup" to the third. Some hosts lock root deliberately and sell "managed" as a synonym for "you cannot see inside". That is a legitimate product, but you should know that is what you bought. Every VICIfast plan ships the native VICIdial admin and full root.

Can you bring your own SIP carrier?

Insist on yes. A host that forces you onto its own SIP is setting your per-minute rate for as long as you stay.

Carrier minutes are usually the largest recurring line in an outbound operation — often larger than the server. When the host resells them, the markup is invisible to you, and two other things quietly become true: you cannot A/B carriers on answer rate and DID reputation, and changing hosts and changing carriers become the same painful project instead of two independent decisions.

Ask:

• Can I point campaigns at my own carrier — Twilio, Telnyx, Bandwidth, a wholesale account?
• Is there any per-minute markup or minimum spend?
• Can I run more than one carrier and route per campaign?

VICIfast is bring-your-own-carrier on every plan with no markup on calls, and the trunk management UI ships templates for the carriers customers most often connect.

Can your team reach the firewall and the agent screen without a ticket?

Agents move. If every IP whitelist change is a support ticket, your floor stops dialing while you wait for a reply.

A VICIdial box is normally firewalled down to known addresses, which is correct — the alternative is a public admin panel getting brute-forced within the day. But that means someone has to add addresses constantly: a home agent on a new connection, a second office, a carrier changing its signalling IP, a new supervisor laptop. The question is whether that person has to be the host.

The agent surface deserves the same question. Can agents log in over HTTPS on a hostname you control, and can they take calls in a browser, or does every new hire need a softphone installed and configured?

VICIfast puts firewall changes in a self-serve portal your supervisors can use, and agents dial from the browser rather than a desk phone or a separately installed softphone. Putting your own logo on that portal and webphone starts at the Scale plan.

How are backups taken, and who can restore one?

Ask for four things: the schedule, the retention window, where the copy lives, and whether you can trigger a restore yourself.

This matters more on VICIdial than on a typical web app. Recordings sit on disk, leads and call logs sit in MySQL, and many of VICIdial's own tables are MyISAM — there is no transactional rollback. A bad bulk operation on a lead list is landed, not undone. Backups are the only undo you have.

Ask:

• How often is a backup taken, and at what local time?
• How long is each one kept?
• Does the copy live on the same box, or somewhere the box's failure cannot reach?
• Can I trigger a restore myself, or do I file a ticket and wait?
• Can I also push a copy to my own storage, which is what you want if dial logs have to survive a discovery request?

VICIfast's Platform Backup takes a daily snapshot at 03:00 server local time on a 7-day rolling retention, with one-click restore, and is available as a $19/mo add-on from the Growth plan up. Pro and above can also send backups to your own S3 or SFTP destination.

How do you get off the box you are on now?

Migration should be the host's job, and it should not require a dialing outage.

This is where host changes actually die. Your current box holds campaigns, lists, dispositions, recordings, agent logins, hot keys, DNC entries and a pile of settings nobody documented. Rebuilding that by hand is a week of work and a week of subtle differences.

Ask:

• Do I land on the same VICIdial version, or an upgrade I have to re-test?
• Do recordings come across, or only the database?
• Does the old box keep dialing while the copy runs?
• Is there a validation pass before cutover, or do I find the gaps in production?
• Who flips DNS, and do I choose when?

VICIfast automates this for ViciBox and self-hosted sources: you point us at the source, the copy runs in the background while the old box keeps dialing, there is a validation pass, and you flip DNS when you are ready. The ViciBox migration path walks through each step, including the carrier IP whitelist change that people forget.

Can you put your own brand on it?

This matters if you resell dialer seats, or if you simply do not want agents and clients to know who your host is.

Whitelabel is not one switch — check each surface separately:

• The login page and the URL agents type. Is it your domain or the host's?
• The agent screen and the webphone. Whose logo is on them?
• The VICIdial admin itself. Does the host's name appear anywhere in it?
• The firewall portal your supervisors use.
• Notification emails.

On VICIfast, login on your own domain is available on every plan except Launch and Scale; the firewall portal and webphone carry your logo from Scale up; and the VICIdial admin is fully white-labeled on Pro and above.

Is the price per agent or per server?

Per-server pricing stops scaling with headcount. Per-agent pricing does not. Do the multiplication before you sign, not after you hire.

Both models are defensible:

• Per-agent is the CCaaS norm. It is easy to expense, easy to forecast per seat, and the vendor's incentive is to help you add seats. The cost curve is linear — doubling the floor doubles the bill, even though the underlying server has not changed.
• Per-server is a flat fee per box, tiered on CPU and RAM, which is what agent concurrency actually consumes. Once you have the tier, adding agents costs nothing until you outgrow the hardware. The risk is sizing: undersize the box and your agents feel it before your accountant does.

Run your own numbers. Take the agent count you expect twelve months out, multiply by the per-agent rate, and compare it against the flat monthly on a box that fits. Then check what is metered on top of either model — per-minute markup, recording storage, DIDs, and support that turns out to be an extra line.

VICIfast is per-server. Plans start at $49/mo, and the shared-CPU tiers step through 5, 10, 25, 50 and 100 simultaneous agents, with dedicated-CPU tiers above that. The full plan grid and what each tier includes is published — no quote required, which is itself worth treating as a criterion.

What does support actually cover?

Get the boundary in writing: what counts as "the server" (their job) and what counts as "your dialer configuration" (usually yours).

Two very different things get called support:

• Infrastructure support — the box is down, Asterisk is not registering, the certificate expired, disk is full.
• Application support — your dial level is dropping calls, an ingroup is not routing, a list is not dialing, a report does not match what the floor says.

Most managed hosts own the first and treat the second as consulting, sometimes on a paid retainer. Some include a slice of it. Neither is wrong. What is wrong is finding the boundary during an outage.

Ask for the hours, the channel, the response expectation, whether anyone will touch your dial plan, and who fixes it when a change you made at root breaks the box.

What happens when you leave?

Ask before you sign, not when you are already unhappy.

• Can I take a full MySQL dump and my recordings, and can I do it myself?
• Is there an export in the dashboard, or do I SSH in and take it?
• What notice do you need, and is the final month prorated?
• How long is my data kept after cancellation, and can I still retrieve it in that window?
• If a payment fails, is the server suspended first or deleted?

A host that answers these plainly is a host that expects to keep you for reasons other than lock-in. Vagueness here is the single most reliable warning sign on this list.

The checklist in one table

Four tests that beat any sales call

Vendors answer questions the way they wish the product worked. These four are cheap and they are answered by the product, not by a person:

1. Sign up and provision a server. Time it yourself. This is the one test nobody can talk their way through.
2. Add your own SSH public key and log in as root. If that needs a ticket, note how long the ticket takes — that is your future recovery time.
3. Invite a coworker. If they cannot get scoped permissions, everyone shares one login, and your audit story is that there isn't one.
4. Ask, in writing, what happens to your data when you cancel. Keep the reply.

Anything a host will not demonstrate before you pay, assume does not exist.

When self-hosting is still the right answer

Managed hosting is not a moral position, it is a staffing decision. Keep the box yourself when:

• You already employ someone who runs Asterisk well, and they would rather own it than escalate to a stranger
• Your compliance regime requires physical custody of the hardware
• You have hardware you cannot move — T1/E1 cards, an on-prem recording appliance
• You run one low-volume box, you are not growing, and an hour of downtime genuinely costs you very little

Outside those cases, the arithmetic usually favours managed, because the recurring hours are real even when nobody invoices you for them.

Where to go from here

Take the ten questions above to two or three vendors and compare the answers side by side rather than the marketing pages. If it helps, we have already done that for the vendors buyers ask us about most, in our head-to-head comparisons.

For our own answers: VICIfast managed VICIdial hosting builds a server in about 40 seconds with root SSH and your own SIP carrier, pricing starts at $49/mo per server with the full plan grid published, and moving an existing floor runs through the automated ViciBox migration. The trial is free for 7 days, which makes test one on the list cost nothing.