We ask for the least access that gets the work done.
Letting an outside team into your systems is a real risk, and it should be priced as one. These are the defaults we work under on every engagement — before you ask, and whether or not it is in the contract.
Least-Privilege, Scoped Access
We ask for the narrowest access that lets the work happen, and we ask for dedicated keys rather than your production credentials. Database access is read-only unless a task explicitly requires otherwise, and write access is granted per task rather than standing. When an engagement ends, every credential issued to us is revoked by you, not by us.
Secrets Never Enter the Repository
API keys, tokens, and connection strings live in environment variables and your secret manager. Nothing sensitive is committed, pasted into an issue, or logged in plain text. Where a build needs a credential at runtime, it reads it from the environment at the point of use — never from a config file checked into version control.
Production Data Is Never the Test Bed
Migrations, destructive operations, and data transformations are developed and verified against a schema copy before they go anywhere near live data. Where a task genuinely requires production access, it is time-boxed, logged, and agreed in advance. We do not run exploratory queries against live customer data.
Security Reviewed as Its Own Workstream
Auth model, permission boundaries, rate limiting, CORS policy, and input validation are scoped and reviewed independently rather than appended at the end of a build. Every service we ship has rate limiting and locked-down CORS from its first deploy, not from the hardening pass that never gets scheduled.
You Own the Code and the Infrastructure
We work inside your GitHub organisation and your cloud accounts from the first commit. There is no handover event because nothing was ever held on our side — no proprietary runtime, no license, no per-seat fee on anything we build. Taking the codebase to another team requires no conversation with us.
Human Approval on Irreversible Actions
For any AI agent or automation we build, actions are classified by reversibility. Anything irreversible — customer-facing sends, payment movement, data deletion, contractual commitments — sits behind a human approval gate. Every tool call is logged with inputs and outputs so any action can be replayed and understood after the fact.
We’ll complete your VSQ before you commit
We are not currently SOC 2 or ISO 27001 certified, and we will not imply otherwise on a questionnaire. What we do provide: a completed vendor security questionnaire, a signed DPA, named subprocessors, and a written description of exactly which systems we would touch and under what access. If certification is a hard procurement requirement, tell us early — we will say so rather than waste your legal team’s time.
Start with the diagnostic
Find out what you should
actually build.
One call to scope it. You leave with an architecture and a costed plan you can take anywhere — including to another vendor.
Or email us directly at hello@techbuilds.io