Integrations
A connection is a list of permitted actions, not a logo.
Every connection on this page is published three ways: the specific operations it may perform, how it is built, and whether you can rely on it today. No product name appears until that connection has been verified end to end.
You will not find a grid of software logos on this page. A logo tells you a name; it does not tell you which operations are permitted, whether the connection has been tested, or whether its owner has agreed to be listed. The table below tells you all three. Product names are published once the connection is verified and the mark is cleared for use.
Two labels, two different questions
Read them together. A Native connection that requires verification for your instance is not the same thing as a Planned one, and neither is a connection you can switch on this afternoon.
Integration type — how it is built
| Type | What it means |
|---|---|
| Native | Built and maintained inside the product. No third-party automation service sits in the path. |
| Via connector | Reached through an automation connector. Adds a dependency you should know about: its uptime and its data path become part of yours. |
| Custom | Built for your deployment against your system’s interface. Scoped and quoted as custom work before it starts. |
| Planned | On the roadmap, not built. Listed so the roadmap is legible. It cannot be selected, cannot be scoped into a pilot, and must not be relied on. |
Status — whether you can rely on it
| Status | What it means |
|---|---|
| Supported | Working today, with the permitted actions exactly as listed. Can be scoped into a pilot. |
| Requires verification | Designed and described, but not yet verified end to end against a real instance. It is agreed in the pilot scope or it is not in the pilot. We will not tell you it works before we have watched it work. |
| Not yet available | Nothing exists to connect to yet. Paired with the Planned type, and never presented as something you could choose. |
Connections, and the exact operations each one performs
The agent has no implicit permissions. If an operation is not on this list for your deployment, it cannot happen — not as a fallback, not as a retry, not as a helpful guess.
| Capability | Permitted actions | Integration type | Status |
|---|---|---|---|
| Reservation and diary systems | Read availability · create a booking · amend an existing booking · cancel on request | Via connector | Requires verification |
| Workshop diary and job booking | Read days open for requests · create a service request · amend a service request · carry an approval status | Via connector | Requires verification |
| Calendar | Read free/busy · create an event · update an event | Via connector | Requires verification |
| Email follow-up | Draft a confirmation · send to the caller · send to a staff address | Native | Supported |
| Customer records | Look up a contact · attach a call note · create a lead | Custom | Requires verification |
| Vehicle records and service history | Look up a vehicle · attach a caller-stated odometer reading · attach a caller-reported symptom, marked as not a diagnosis | Custom | Requires verification |
| Staff notification channel | Post a follow-up request to a named channel · post an unknown-outcome alert asking for a record to be checked | Custom | Requires verification |
| Telephony and inbound numbers | Receive an inbound call on a number you control · end the call | Via connector | Requires verification |
| Live transfer to a person | Warm transfer · request a callback · notify a staff channel | Planned | Not yet available |
How a connection moves from Requires verification to Supported
The same three stages every time, and the status on this page only changes at the end of them.
-
The operations are written down
Before any code, the exact list: which reads, which writes, which fields, and what the agent does when the system answers with an error or does not answer at all.
-
It is run against a real instance
Not a mock. A real account of that system, exercising each permitted action and each failure path, including the case where a write is accepted but not confirmed.
-
The status and the actions are published
The row above changes to Supported and the permitted actions become the list we will be held to. If the vendor’s mark is cleared for use, the product name is published with it.
What happens when a connection fails mid-call
- A read that fails is said out loud. If availability cannot be read, the agent says it cannot check rather than offering a time it has not seen.
- A write that is not confirmed is never a tick. The record of what Marlin did records an not confirmed and asks for the record to be checked. No silent retry, and no green tick the system cannot prove.
- A connector outage is not the caller’s problem. The call falls back to capturing the request and routing it to a person, which is the behaviour a connection-less deployment has anyway.
Questions about connections
Do you connect to the system I already use?
Tell us what it is in your pilot request and we will answer specifically. If it falls into a category above with a connector, it is setup work. If it has no connector, it is custom work and is quoted before it starts. If neither is true we will say so rather than putting your product name on this page.
Why does Email follow-up say Supported when everything else does not?
Because drafting and sending an email is inside the capability the team has actually built and tested. Writing into a third-party booking record is not yet verified for any deployment. Publishing both as “integrations” in the same visual style would hide the difference that matters most.
Can I choose a Planned connection for my pilot?
No. Planned means nothing exists to connect to. It is listed so you can see where the product is going and can tell us if it is a blocker, but it is not selectable and it cannot be scoped into a pilot agreement.
What can the agent do with no connections at all?
Answer the call, hold the conversation, capture the request with the details your workflow needs, and send the follow-up email. What it cannot do is write into a system it has no permission to reach. For some businesses that is already the useful half; for others it is not, and a pilot is the way to find out which you are.
Who can see the data that moves through a connection?
That depends on the connector in the path and on your own system’s permissions. What is logged, who can see it and how to request a deletion are set out on data handling and controls, including the items that are still pending legal confirmation.
Tell us which system has to be in the loop.
Name it in your pilot request. We will tell you whether it is setup work, custom work, or not possible yet — before anyone signs anything.