Cover photo: Albert Stoynov / Unsplash. Illustrative photograph, not a record of a client engagement.
A software agency handover should transfer the ability to run and change the product, not just deliver a ZIP file of source code. Inventory ownership and access, agree the transfer steps, and have the receiving team demonstrate a build, deployment and recovery using the handed-over material.
This checklist is for a founder who has already decided to end or change a development-supplier arrangement. It is not an investment due-diligence exercise or an enterprise go-live plan. The question is whether the product remains under your control when the supplier leaves.
Agree the exit before moving accounts
Name an outgoing lead, a receiving technical owner and a business acceptance owner. Agree the transition period, support arrangements, deliverables and how disputed items will be handled. Check the contract with an appropriate adviser where ownership, notice, licences or payment obligations are unclear. Paying for development is not enough evidence to resolve every intellectual-property question.
The UK DDaT Playbook’s exit-planning chapter supports early planning, knowledge transfer and continuity. The Playbook does not by itself impose obligations on every private UK or UAE business; check requirements incorporated into your contract. The operational checklist here is a suggested method, not a legal interpretation of your contract.
If the relationship is strained, keep the checklist factual. Record the item requested, contractual position if established, due date and evidence received. Do not attempt to take access you are not authorised to use.
Build an ownership inventory that goes beyond Git
For each asset, record the legal or contractual owner, current administrator, billing owner, receiving owner, transfer method and acceptance evidence. Mark disputed or unknown ownership explicitly.
Include the repositories and their history, branches, tags, large-file storage and package registries. Then follow the product outward: cloud accounts, databases, object storage, domain registrar, DNS, certificates, email delivery, monitoring, analytics, app stores and other external integrations. List the build pipeline, runners, scheduled jobs and infrastructure definitions too.
A repository in your organisation can still depend on a package in the agency’s private registry. A cloud account can be yours while its recovery email belongs to somebody who has left. Those are dependencies to resolve, not footnotes.
Check third-party components and licences separately. Some supplier tools may be licensed rather than transferable. Record an authorised replacement or continuing arrangement where a transfer is not possible. Keep payment-card details, passwords, tokens and recovery codes out of the worksheet.
Transfer authorised access without breaking the service
Use each provider’s supported ownership and administrator-transfer process. Establish named receiving accounts and verify recovery routes. Share secrets only through an approved secure mechanism; a spreadsheet of passwords is not a handover pack.
Sequence changes with the people operating the service. Identify which applications use each credential before rotating it, update the dependent systems and verify their operation. Keep a time-limited support route during the agreed transition. Revoking everything first can remove the only working recovery path.
Record which old identities, deployment keys, integrations and sessions need to be removed, who is authorised to remove them and when. Verify the replacement access before removal. Do not assume deleting a person’s account also removes every machine credential they created.
For teams across locations, name who can approve account changes and who can reach provider support during the agreed working window. Authority and availability matter more than the location on the contract.
Ask the receiving team to demonstrate control
The outgoing agency can explain and observe. The receiving team should operate the controls. Use an isolated, authorised environment with synthetic or appropriately protected data, and agree the scope so the exercise cannot accidentally send customer messages or take live payments.
- Build: start from a clean environment, obtain the specified code version, install the documented dependencies and produce the artefact. Capture the version and build log. Record anything fetched from an undocumented private source.
- Deploy: use the handed-over pipeline or instructions to deploy to the test environment. Confirm configuration, background jobs, logs and a representative user journey. Demonstrate the documented rollback there.
- Restore: recover a suitable backup into an isolated destination, check that the application can use it and record the backup timestamp and observed recovery duration. Compare the result with the business’s agreed recovery requirements; do not invent a universal acceptable time.
- Diagnose: give the receiving team a controlled failure or a historical incident scenario. Can they find the logs, identify the owner and follow the support instructions?
A successful build does not prove recovery. A restore that produces a database file but no usable application is incomplete evidence. Record each result separately.
Leave defects and commercial exceptions visible
Illustrative example: the new engineer can build the product but cannot deploy it because a private runner belongs to the agency. Record the runner as a blocked control-transfer item. Agree either an authorised transfer or a replacement runner, then repeat the deployment test. Do not mark the handover accepted because the repository arrived on time.
Use the worksheet to separate asset transfer from acceptance. It includes prompts for build, deployment, recovery, support and access removal, plus blank evidence and reviewer fields. Statuses can be not started, in progress, evidenced, accepted or blocked. Each exception needs its impact, owner, action and deadline.
List open defects, warranties or support commitments, known vulnerabilities and upcoming renewals. Confirm who will respond after the exit date and what help remains chargeable. Do not silently turn technical acceptance into a waiver of contractual rights; get the relevant review where that distinction matters.
The business acceptance owner should record the decision: accepted, accepted with named exceptions, or not accepted. A critical dependency on the outgoing supplier should be visible in that decision. A folder of documents is a delivery. Independent operation is the evidence.
Practical questions
Is a repository export enough to accept the handover?
No. It may omit dependencies, account control, configuration and operating knowledge. Accept against the agreed inventory and demonstrations, with exceptions recorded individually.
What if an account cannot be transferred?
Ask the provider and supplier about the supported options. Agree a replacement or continuing arrangement, including ownership, cost and access, and test the resulting dependency. Do not try to bypass an account restriction.
Should agency access be removed on the last contract day?
Agree the timing against verified replacement access and any authorised transition support. Record each identity and machine credential separately. A calendar date alone does not prove that revocation is safe or that ongoing access is justified.
Can the founder accept without a receiving technical team?
The founder can own the business decision, but needs someone competent to inspect the evidence and perform the demonstrations. Arrange that capability before the exit date rather than relying entirely on the outgoing agency’s assurance.

