Handing a Cloud Account Over to a Client Cleanly
Transferring an account you built to the client who paid for it. Covers scoping the handover, separating your tooling, switching billing, and what access you keep afterwards.
Most guidance about account transfers is written for the person receiving the account. This one is written for the person giving it away. The two jobs are not mirror images: the outgoing operator knows everything and has to decide what to pass on, while the incoming one knows nothing and has to decide what to trust.
Done badly, a handover leaves your credentials in a system you no longer monitor, your client unable to answer a provider’s questions, and both of you unclear about who pays next month’s invoice. Done well, it takes an afternoon and ends with a short document. The difference is almost entirely in the preparation.
Decide what is actually being handed over
Before any technical work, settle the scope in writing. There are usually three candidates and they are routinely confused. The first is the cloud account itself, including its billing relationship. The second is the infrastructure inside it. The third is the operational responsibility for keeping that infrastructure running.
A client can take the account without taking the operations, and frequently should. A client can also take operations while you keep an advisory role. What causes trouble is leaving the question implicit, so that in week three nobody is watching the alerts because each party assumed the other was. Write one sentence per item stating who holds it after the handover date, and have both sides agree to it.
Also settle who the registered account holder will be. If the account was created in your company’s name, the client may want it recreated in theirs, which is a different and larger job than transferring access. Establish that early, because it changes the plan entirely.
Clean the account before you open it
Nobody wants their client’s first look at an account to include a half finished experiment from eight months ago. Spend an hour on cleanup before the handover, and do it properly rather than cosmetically.
Delete resources nobody uses: stopped instances, orphaned volumes, old snapshots, unused load balancers, empty buckets, and test projects. Remove policies that no longer attach to anything. Consolidate the naming so related things look related. Check every region, not just the one you work in, since the forgotten region is a standard feature of agency accounts.
Then look at data. Anything belonging to another client has no business being in this account, and discovering it after the handover is a conversation you do not want to have. This is the moment where the argument in the case for account isolation pays off: if you kept one account per client from the start, this step takes minutes instead of a day.
Separate your tooling from theirs
Agency accounts accumulate tooling that belongs to the agency rather than the client. Monitoring agents reporting to your dashboard, backup jobs writing to your storage, CI runners authenticating with your credentials, cost dashboards pulling into your reporting. All of it needs a decision before the account moves.
For each integration, choose one of three outcomes. Remove it, if it was only ever for your convenience. Hand it over, if the client is taking it on, which means they need their own account with that vendor rather than a seat on yours. Or keep it running under an explicit agreement, with its own identity in the account and a documented purpose, so that the client’s next security review does not flag it as an intrusion.
Pay particular attention to anything that writes outside the account. A backup destination you control is a dependency the client cannot see, and it will fail silently the day you reorganise your own storage. Move the destination into the client’s account before the handover, not after.
The billing switch
Billing is where handovers most often stall, because it involves a payment method, a legal entity, and sometimes a tax registration. Plan it as its own step with its own date.
Work out in advance whether the provider lets you change the billing profile on an existing account or whether it requires a new one. Find out whether the billing country can be changed, since on many platforms it cannot, and if the client needs invoices from a different jurisdiction then a new account is the only route. Confirm whether any committed spend agreements, reserved capacity or discount arrangements survive the change, because those are contractual rather than technical.
Agree who pays for the overlap period. A handover in the middle of a billing cycle produces one invoice covering both of you, and the cleanest resolution is usually for the outgoing party to pay that cycle and the incoming party to start fresh at the next one. Decide it beforehand rather than when the invoice lands.
One more billing detail that catches people: alerts and budgets are usually tied to the billing profile rather than to the account, so changing the profile can silently orphan every threshold you configured. Recreate them under the new arrangement and confirm one fires, rather than assuming they carried across.
Running the handover session
Do the credential transfer live, with both parties present, rather than by email. The order matters and it is the reverse of the order the receiving party would choose.
Start with the registered email address, since everything else flows from it. Either transfer control of the mailbox or change the account’s address to one the client owns, completing any confirmation step while you are both available. Then have the client create their own owner level identity and verify they can sign in with it. Then have them enrol their own multi-factor device. Only at that point do you remove your own factors and credentials.
Finish by having the client perform one meaningful administrative action themselves while you watch, such as creating a user or changing a budget. It confirms their access is real rather than theoretical. The receiving party’s follow up work is set out in securing an inherited account, and it is reasonable to send them that before the session so they know what comes next.
Documentation the client will read
Long handover documents are a way of transferring blame rather than knowledge. Write a short one. It needs the account identifier and login URL, the list of what runs in the account and roughly what each thing does, the regions in use, where logs and backups go, the names of any remaining integrations and why they exist, and a list of known issues or deferred work.
Add two things people usually omit. The first is a map of external dependencies: domain registrar, DNS provider, certificate issuance, third party APIs with their own credentials. The second is what to do if the provider suspends the account, including which contact address receives the notice. If the infrastructure is defined as code, point at the repository and say which parts are live, since stale infrastructure code is worse than none at all. The provider model in infrastructure as code is useful background if the client is inheriting a codebase as well as an account.
Your access after the handover
Default to having none. If the engagement has ended, remove your identities, revoke your keys, and tell the client in writing that you have done so and on what date. That statement protects you as much as it protects them, because it establishes when your responsibility stopped.
If you are staying on in a support capacity, do not keep the owner credential. Create a named identity with the permissions the support role actually needs, make it subject to the same review as anyone else’s, and agree an expiry date for it. Where the provider supports time limited or just in time access, use it. Shared logins between an agency and a client are a liability for both sides and they are never worth the convenience.
The thirty day tail
Expect loose ends. Certificates renew, scheduled jobs fire monthly, backups rotate, and an annual domain renewal can surface weeks later. Agree a short support window, typically a few weeks, during which you will answer questions about things you built, and be specific that it is advisory rather than operational.
Use that window to check one thing yourself if the client permits it: that the billing alarm fires, that the backups are still being written, and that the audit log is still recording. Those three are the ones that fail quietly after a transfer. After that, close the file. If you are setting up a fresh account for the next engagement rather than reusing anything, the account catalog and provider pages such as Buy Digital Ocean Account and Buy Hetzner Account are the usual starting points, and DigitalOcean against Linode covers the trade-off most agencies weigh first.