Can a connected agent send messages or spend money?
Download MarkdownIt can never send a message: no text, no email, and no call to a lead, a seller, a buyer, or a title company. It can spend money in exactly one way, and only if you grant it that scope: requesting a comp report, which is metered p
It can never send a message: no text, no email, and no call to a lead, a seller, a buyer, or a title company. It can spend money in exactly one way, and only if you grant it that scope: requesting a comp report, which is metered per run.
Most of the tools only read. A small number write, and none of them sends a message to a lead, a seller, a buyer, or a title company (a stage change gives your own deal team the usual internal notice, described below). Two of the writes prepare messages and stop there: one drafts a single follow-up to one lead, the other assembles a blast campaign with its recipients already queued and suppression applied. Both come back at the status "draft". Nothing sends, schedules, scrubs a number, or spends a credit until a person approves it in the app, and the approval and launch calls are not reachable from the credential an agent authenticates with.
The remaining writes change your own records. They put an offer on one of your properties or amend its commercial terms, mark a task complete on a transaction workfile or a fix-and-flip project, and append an internal note to a transaction's communication log or a flip's note log. They also create or update a property, set a property's deal owners (acquisitions, dispositions, follow-up, and transaction coordinator), move a property's stage with the assigned buyer and the assignment fee, reopen a transaction task, mark one not applicable, or reassign it, edit the fields of a transaction file, and add or update the people listed on it. Notes are records, never outbound messages.
Offer decisions and EMD received stay with a person. Accepting, declining, and backing up an offer, marking an offer's EMD received, attaching a contract, and deleting anything all stay with a person in the dashboard, and an offer a person has already decided is frozen. The stages from Buyer EMD through Closed also need an accepted offer with EMD received, so an agent's move to one of them is refused until a person has recorded it. An agent can write the EMD amounts, holders, and dates on a transaction file, but those do not satisfy that rule.
Changing a stage is the one agent action that reaches your LeadAttractor CRM, and it does what the same change does by hand. When a deal moves to certain stages, CloseMagnet adds an internal note to the seller's contact in your LeadAttractor CRM (for example "Offer accepted", or "Deal fell through" with the reason), provided the property has a seller contact linked. When a deal closes or is lost, it also adds the closed-deal or lost-deal tag. Your own deal team gets the same internal notification email a hand change sends. The note and the tag are internal records, and CloseMagnet does not contact the seller. If you have an automation in your LeadAttractor CRM that reacts to one of those tags, it reacts the same way it does to a hand change. If the note or tag cannot reach your LeadAttractor CRM, the stage change still stands, exactly as it would by hand, and the answer tells the agent so. Nothing else an agent does is sent to your LeadAttractor CRM, and your deal pipeline is not synced to it.
Every property and transaction-file write can run as a dry run first. CloseMagnet checks the whole change and shows what would change, and it saves nothing, adds nothing to your LeadAttractor CRM, and emails no one. A call repeated with the same idempotency key returns the first answer and changes nothing twice.
The one write that costs money is the comp report. After your plan's free comps for the month, each comp is charged to your CloseMagnet balance, so the call is refused unless it explicitly confirms the spend, and it is refused outright unless the credential carries the comps write scope. So if you do not want an agent spending, leave Write clear on the Comps row when you mint its credential. In the same way, if you do not want an agent changing properties or transaction files, leave Write clear on the Properties and Transactions rows. An agent cannot perform a billable directory reveal at all; that happens in the app.
If an agent tells you it sent a message, it is wrong by construction. Go look at the draft.
Related Articles
- What can the CloseMagnet API do?It reads almost everything in your account, and what it writes is deliberately narrow. Its endpoints span 17 domains and the large majority are read-only. The ones that write do one of these things:
- How do I create an API credential?Only the account owner can mint one, and it happens at Settings -> API.
- What is the difference between my account API token and a scoped credential?Your account API token was minted automatically when the account was created. It sits pinned as the first row in Settings -> API with the type "Automatic", carries a fixed set of scopes, can be revealed again at any time, and can
- How do I rotate or revoke a credential?It depends which kind you have, and only the account owner can do either one.