Start typing to search this publication.
techcirkle
Open menu

Subscribe to techcirkle

Get new posts delivered straight to your inbox.

The Contract Clauses That Decide Who Controls Your Codebase

IP assignment, repo location and infrastructure ownership decide who has leverage when a vendor relationship ends. Most engineers never see these clauses.

techcirkle
6 min read
Cover image for The Contract Clauses That Decide Who Controls Your Codebase

Most engineers never read the master services agreement. It is not our document, it arrives after the technical evaluation, and it is written in a register specifically designed to be skimmed. Then eighteen months later somebody wants to change vendors, and it turns out the answer to "can we?" was written down at kickoff by people who were not thinking about this moment.

Four clauses do most of the work. They are all technical in effect even though they are legal in form, which is why the engineer in the room is often the only person who understands their consequences.

1. When IP assignment takes effect

The clause you want says intellectual property in work product assigns to the client on creation, unconditionally.

The clause you often get says assignment occurs upon receipt of final payment.

The second version sounds like a normal commercial protection and functions as something quite different. It means that during any billing dispute — including a good-faith one about scope — the vendor holds legal title to code that is running in your production environment. The leverage in that situation runs entirely one direction.

It is a one-line change and it is not usually contentious once raised. It just has to be raised, because it will not be volunteered.

Related and easier to miss: check whether assignment covers everything or only "deliverables." If deliverables are defined as the application, that may exclude infrastructure-as-code, CI configuration, internal tooling, migration scripts, and documentation. Those are exactly the artifacts a new team needs most, and exactly the ones a departing vendor has least incentive to hand over cleanly.

2. Where the repositories live

This one is a day-one operational decision with contractual consequences.

The correct arrangement: repositories in your organisation from the first commit, with vendor engineers as members. Not "the vendor develops in their org and transfers at handover."

The transfer version fails in predictable ways. Commit history gets flattened or lost. Issues, pull request discussions, and the reasoning behind decisions live in the vendor's tenancy and do not come across. Branch protection rules, CI secrets, and deployment configuration are rebuilt from memory. And the handover becomes an event with a negotiating position attached, rather than a non-event.

Arranging it correctly at kickoff costs one conversation. Unwinding it later costs weeks and, occasionally, money.

3. Who owns the cloud accounts

Same principle, higher stakes.

Cloud accounts should be owned by your company, with vendor engineers holding scoped IAM roles. The inverse — vendor-owned account, client as a guest — puts your production environment, your data, your DNS, and your billing relationship inside somebody else's tenancy.

Ask specifically about:

  • The cloud account root credentials and who holds them

  • Domain registrar access

  • CI/CD platform ownership and secret storage

  • Monitoring and log aggregation accounts

  • Third-party service accounts — payment processors, email providers, error tracking, feature flags

That last category is the one that gets forgotten and then discovered at the worst possible moment. A Stripe account or a Sentry organisation registered to the vendor's email domain is a small thing until it is not.

4. Proprietary frameworks and background IP

Many firms build on internal accelerators — a boilerplate, an auth module, an admin scaffold, a deployment toolkit. This is legitimate and often genuinely valuable. The problem is when it is not disclosed.

The clause to look for grants you a licence to use the vendor's background IP within the delivered product, while ownership stays with them. That can be perfectly acceptable — but only if you know which parts of your codebase it covers, and only if the licence is perpetual, irrevocable, transferable on acquisition, and does not require ongoing payment.

Where this hurts is due diligence. An acquirer's technical review asks what proportion of the codebase you own outright. If the answer involves a vendor licence nobody documented, the deal gets slower and the price gets discussed again.

Ask for the list up front. A firm that is comfortable with its own model will produce it without friction.

The exit clause, which is really a technical spec

Buried near the end of most agreements is a termination and transition section, usually one paragraph of boilerplate. It is worth rewriting as a technical requirement.

What good looks like:

  • A defined transition window with a stated notice period

  • Documented handover artifacts specified concretely — architecture documentation, runbooks, ADRs, deployment procedures, credential inventory, and known-issue list

  • Knowledge transfer sessions at agreed rates rather than at whatever rate is quoted under pressure

  • Explicit confirmation that no production dependency runs on vendor-controlled infrastructure

That last point deserves a check rather than an assurance. It is common for a single scheduled job, a monitoring integration, or a build agent to be running somewhere in the vendor's environment simply because it was convenient at the time. It is discovered when it stops.

Why this lands on engineers

Because everyone else in the room is evaluating the clauses as commercial risk, and their consequences are almost entirely technical.

A procurement lead reading "assignment upon final payment" sees a payment term. An engineer sees a scenario where the running system belongs to someone else during a dispute. A finance lead reading about a vendor-owned cloud account sees an operational convenience. An engineer sees a migration project with an unknown length.

You do not need to be the decision maker. You need to read four clauses before signature and say what each one means in practice. That is a twenty-minute contribution with a multi-year effect, and it is one of the highest-leverage things a senior engineer can do outside of code.


The full buyer-side guide — delivery models, US rate bands, evaluation criteria, compliance expectations, and realistic twelve-month engagement economics — is here: Product Engineering Services Companies in USA: A 2026 Buyer's Guide.

Frequently Asked Questions

Is "assignment on final payment" ever reasonable?

It is a normal commercial instinct, but the consequence is disproportionate: title to code running in your production environment sits with the vendor during any dispute. Assignment on creation with normal payment remedies elsewhere in the contract achieves the vendor's legitimate protection without that side effect.

What if the vendor insists on developing in their own GitHub org?

Ask why. Occasionally there is a real reason involving shared internal tooling, in which case agree an explicit mirroring arrangement. More often it is habit. The cost of changing it at kickoff is a single conversation; the cost of migrating later includes lost issue history and rebuilt CI configuration.

Does IP assignment cover infrastructure-as-code and CI config?

Only if the definition of work product is broad. If assignment is limited to "deliverables" and deliverables means the application, then Terraform, pipeline definitions, runbooks, and internal tooling may fall outside it. Those are exactly what an incoming team needs most.

Are vendor proprietary frameworks a dealbreaker?

No, provided they are disclosed, scoped, and licensed perpetually, irrevocably, and transferably on acquisition without ongoing fees. The problem is undisclosed background IP surfacing during acquisition due diligence, where it slows the deal and reopens the price.

What should a transition clause actually specify?

A notice period, a defined transition window, a concrete artifact list (architecture docs, runbooks, ADRs, deployment procedures, credential inventory, known issues), knowledge transfer at pre-agreed rates, and confirmation that nothing in production depends on vendor-controlled infrastructure.

Should engineers really be reading contracts?

You do not need to read the whole thing. Read the IP assignment clause, the infrastructure and access provisions, the background IP section, and the termination clause — then translate each into its practical consequence for the people deciding. Nobody else in the room can do that translation.

Subscribe to techcirkle