Your code stays where you put it
Testver was built local-first, so most security questions have a short answer: it runs on your hardware, your tests live in your repository, and you decide what the AI is allowed to touch.
It runs on your machine
Testver installs as a normal package and serves its UI from localhost on your own hardware — or on a server your team controls. There is no Testver-hosted environment executing your tests.
Your code stays your code
Tests are written into your repository as plain files in your framework. Nothing is uploaded for storage, nothing is kept behind a login, and removing Testver leaves the suite untouched.
Every tool call is gated
The assistant cannot read a file, run a command, or reach an external system without an approval step. Permissions are per-action and visible, including for MCP tools you connect yourself.
Offline is a supported mode
Point Testver at a local model through Ollama or LM Studio and the whole workflow runs with no outbound internet traffic at all — the same features, nothing leaving the building.
What leaves your machine — and what never does
When you use Testver’s cloud AI, this is the complete list. On a local model, the left column is empty.
The prompt you typeSent to the model
Sent to the model you selected, so it can answer.
Code snippets the task needsSent to the model
Only the files or fragments relevant to the request — chosen for the task, not the whole repo.
Failure context when you ask for a fixSent to the model
The error, the relevant trace or DOM fragment, and the failing locator.
Your full repositoryNever sent
Never bulk-uploaded or mirrored anywhere.
Your test results and reportsNever sent
Stored locally with your project. Testver has no copy.
Your credentials and secretsNever sent
Environment variables and secrets stay in your environment.
Anything at all, on a local modelNever sent
With Ollama or LM Studio configured, there is no outbound request to make.
The same rule applies to MCP servers you connect: they run where you run them, and their tools are approved exactly like the built-in ones.
The questions a review usually asks
Deployment
Install per-developer on a workstation, or run one shared instance on a machine inside your network. Both are supported and neither requires opening inbound ports to the internet.
Model choice
Use Testver’s cloud AI, or connect a locally hosted model. The choice is per project, so a sensitive repository can run on a local model while the rest use the cloud.
Auditability
Tool calls are visible as they happen, and changes the AI proposes arrive as reviewable diffs you apply or discard — including Auto-Heal fixes, which can be undone after the fact.
Account data
The only thing tied to your Testver account is billing and AI credits. Test code, results and project data are never part of it.
Nothing leaves the building
For regulated work, run Testver on an internal machine and point it at a local model. You keep the recorder, the AI assistance, Auto-Heal, scheduling, reporting and CI integration — with no outbound request in the loop.
- Local model via Ollama or LM Studio
- No account needed to run tests
- Test data and reports stay on disk
- Per-project — mix local and cloud
Have a security questionnaire to fill in? Send it over — we answer them directly. See also our privacy policy.
Keep your framework. Add AI and a premium toolset.
Set up in minutes. Self-host on your own hardware. Describe a test in plain English and get working code in your own framework — in a fraction of the time.