Use case

An API backendwith no screen.

With no page to look at, mistakes hide until a client sends the wrong thing. Adams adds the checks a screen would give you.

Lite and Pro
// the problem

What goes wrong

Nothing looks wrongA wrong response is still a response. The first sign of a bug is often a customer integration that fails.
Input is untrusted by defaultEvery field a client sends can be missing, huge or hostile. AI code tends to trust it.
// what adams does

What Adams does

Guards secrets and gitLite: keys, tokens and risky git commands are held before they do harm.
Reviews with a second agentPro: the reviewer ranks problems such as missing validation and weak error handling.
Scopes security by areaPro: check input handling and access on the endpoints you name.
// worked example

An example

Reviewing an endpoint

Illustrative session.

~/project~/your-project · example
/adams:secure
Scope: input validation
P1: POST /orders accepts any quantity
P2: errors return internal messages
Fix these findings? Yes (recommended)
// faq

Questions

Does Adams test my API?

It does not send requests. It reads the code and runs your own checks and tests when you ask.

Which languages?

Adams works on whatever your AI writes. Python and Node have their own pages here.

Which part is free?

Secret, git and leak guards.

// related

Keep reading