Useful Work, Before You Sign Up.
Every agent here is published in full before anybody asks you for an email address: the job, who it is for, what it reads, what it produces, and the step it stops at and asks you about. Browsing needs no account. Running one needs a Tools account, which is a separate sign-in on a separate platform.
The five jobs below are concepts. They are specified and not built, and none of them can be subscribed to. The prices further down are what a published agent will cost, not something you can buy today.
Available Tools.
Read from the live Tools catalogue each time this page loads, rather than from a list somebody keeps up to date by hand.
Nothing is published yet
The Tools catalogue answered, and it is empty. No agent is running and none can be subscribed to today. The plan prices below are what a published agent will cost, not something you can buy now.
A Tools account is a separate account.
The Tools application runs on its own platform, with its own database and its own sign-in. A voyantiq.com account — the one that buys an application — is not a Tools workspace account, and a session here cannot reach anything there. This page said otherwise until recently and it was wrong.
Nothing about that changes what you can read. The catalogue, every detail page, the prices and the limits are public. The separate sign-in is what you make when there is finally something to run, and it is made over there rather than here.
In Development.
Five jobs, specified in full and not built. Each has its own page with the run laid out step by step, the shape of what it would produce, and what it is not. Every example on those pages is illustrative — nothing here has run, so no figure on this site is a measurement.
Invoice Matching
FinanceChecks every invoice against the order it came from and the goods that actually arrived.
- For
- An accounts payable clerk or a bookkeeper opening every invoice by hand against an order and a receipt.
- Reads
- Invoice documents, Purchase order export, Goods receipt records
- Produces
- A queue of matched invoices, and an exceptions list naming what disagreed on each one it held.
In DevelopmentNot purchasable. No price, and no subscribe button until it runs.
Stops and asks: Approve the 39 and hold the 3?
Month-End Close
FinanceClears the accounts that reconcile at month end and writes the first draft of the close narrative.
- For
- A financial controller or an accountant running the same close in the same order every month.
- Reads
- General ledger export, Sub-ledger exports, Last month's close file
- Produces
- A set of cleared accounts, the differences it could not explain, and a draft narrative waiting for somebody to post it.
In DevelopmentNot purchasable. No price, and no subscribe button until it runs.
Stops and asks: Post the narrative to the close file?
Incident Triage
EngineeringTurns a night of alerts into a short list of grouped incidents, each with somewhere to look first.
- For
- Whoever is on call at two in the morning, and whoever has to write the summary afterwards.
- Reads
- Alert feed, Deploy log, Status history
- Produces
- Grouped incidents, each naming the service involved, carrying a drafted impact summary and saying what is still unknown.
In DevelopmentNot purchasable. No price, and no subscribe button until it runs.
Stops and asks: Page the owner on call?
Review Before Merge
EngineeringReads a change the way a careful colleague would, before anybody is asked to approve it.
- For
- A developer waiting on a review, and the reviewer who has fourteen files to read at five in the afternoon.
- Reads
- The diff, Your written standards, Test history
- Produces
- A drafted review naming each finding with the file and the line it belongs to, and naming the files it had nothing to say about.
In DevelopmentNot purchasable. No price, and no subscribe button until it runs.
Stops and asks: Post the review?
Idle Cloud Spend
EngineeringFlags cloud resources with no recorded activity, works out who owns each one, and estimates what stopping it would save.
- For
- Whoever gets asked why the cloud bill went up, and has to find out from three exports that do not agree.
- Reads
- Billing export, Tag inventory, Owner directory
- Produces
- A list of flagged resources, each with an owner against it and an estimate of what stopping it would save, with the assumptions behind the estimate printed beside it.
In DevelopmentNot purchasable. No price, and no subscribe button until it runs.
Stops and asks: Open a ticket for each owner?
Built, and not published yet.
One package exists in the Tools repository beyond the concepts above. It is built and it is not in the live catalogue, which is a third state and not the same as either of the other two. It is named here because leaving it unnamed is how it got confused with one of the concepts.
Bill Difference Checker
Compares exactly two versions of the same supplier bill and names every line that changed between them.
- For
- A business checking a recurring supplier bill against last period's, line by line.
- Produces
- Charges added, charges removed, charges that changed, the total difference, and the lines it could not read with certainty.
- Availability
- Built in the Tools repository and not published to the live catalogue, so it cannot be subscribed to and carries no price here yet.
- Not to be confused with
- Invoice Matching, which is a concept and a different job: reconciling an invoice against a purchase order and a goods receipt. A bill comparison is not a purchase-order reconciliation.
What it cannot do
- Two bills at a time, both text-readable. A photograph of a bill is not supported: there is no optical character recognition in it.
- It compares the two documents you give it. It does not connect to a supplier account and it does not fetch anything.
- No supplier negotiation, no fair-price benchmark, and no twelve-month trend. It answers one question: what changed between these two.
- Where a line cannot be read with certainty it is reported as uncertain rather than guessed at.
What a published agent costs.
Two plans, each priced per agent per workspace. Neither can be bought today, and both are on the page anyway, because a price you only learn after signing up is not a price.
Standard Agent
$29USD a month
per agent, per workspace
One agent doing one repeated job against sources you connect yourself.
What the month includes
An allowance measured in documents processed.
The size of the allowance is still being set. Real model, interface, retry, infrastructure and support cost is being measured against actual runs first, so no number of runs is published here and no unlimited use is promised. Checkout stays closed until that number exists.
Can it be bought today
Not open yet. Billing and the first published agent are both unbuilt, and the allowance has not been measured, so nothing on this site can take a subscription today.
Advanced Agent
$79USD a month
per agent, per workspace
An agent whose work is heavier per item or spans several connected sources at once.
What the month includes
An allowance measured in reports completed.
The size of the allowance is still being set. Real model, interface, retry, infrastructure and support cost is being measured against actual runs first, so no number of runs is published here and no unlimited use is promised. Checkout stays closed until that number exists.
Can it be bought today
Not open yet, and this tier is only ever attached to an agent whose workload genuinely needs it. An empty tier put on a page to fill a table is not an offer.
How the allowance is counted
The same rules on both plans. They differ in price and in the size of the allowance, never in how a unit is counted.
- Each published agent names its own unit — documents processed, reports completed — on its own page, before you subscribe.
- A unit is counted once when the agent finishes work on it, whether the result is a clean pass or an exception held back for you.
- A retry the agent starts by itself is part of the same unit and is never counted again.
- A run that stops to ask your permission is not counted a second time when you answer.
- A run you cancel before the agent reads anything is not counted at all.
- Resubmitting the same item after you change it is new work and is counted again.
- When the allowance is used up the agent pauses and waits. Nothing is charged beyond the plan price unless you buy a top-up on purpose.
A team bundle has been discussed and is not offered. It has no defined contents, permissions, allowance or operating cost yet, so there is nothing here to price.
What counts as a tool here.
Not every useful thing qualifies. These are the rules a tool has to obey before it goes in the catalogue, and they are the reason the catalogue is short.
It does one job.
And it is a job somebody was doing by hand on a Friday afternoon.
It does it the same way every time.
And the run is recorded, so you can see what it looked at.
It asks permission before a consequence, not before a calculation.
Reading, checking and comparing change nothing outside your own workspace and happen on their own. Sending, paying, posting and deleting stop and wait for you. A read-only agent may have nothing to approve at all, and giving it an approval button would be theatre rather than control.
It works from your own connected accounts.
In a Tools workspace no other customer can see, on the Tools platform rather than on this site.
You cancel and it runs to the end of the cycle you paid for.
Access ends at the paid-through date rather than the moment you click cancel, and you can export what is yours before and after. Nothing is held back to make leaving difficult.
What is published, and what never is.
The boundary is inside each agent rather than in front of all of them. Everything about the work is public. Nothing about the judgement is, and that holds in this site’s markup, its data and its structured output, not only in what a page displays.
What each one does, in full.
The job, who it is for, the kinds of source it reads, the run step by step, what it produces, the step it stops at, and the things it is not. All of it is on this site, without an account.
What it will not do without you.
Stated on every card and every page, because a promise made only after you have paid is not worth making.
Never how it decides.
The instructions it runs on, the lines it draws and where, the cases it has to pass before it ships, and what it falls back to when something fails. That is the part that took the time, and none of it is on a public page, in a comment, in this site's data, or in anything it sends a crawler.
The honest state of this, in four sentences.
Five concepts are specified and unbuilt. One package is built and unpublished. The live catalogue publishes nothing today. Billing does not exist yet, and the usage allowance has not been measured, so there is no checkout on this site to put a card into.
What is real is the platform underneath: its own identity realm, its own tenant database, and a catalogue console whose operator role is deliberately separate from workspace membership. Two workspaces cannot see each other, and that is proved against a real database rather than asserted.