It is tempting to open a JavaScript protector and immediately pick the strongest-looking preset. That habit usually comes from treating protection as a badge: if High exists, Low must be a compromise you should outgrow. MeowShield's own homepage says the opposite. Start with the smallest Low, Medium, or High level that meets the threat model, then verify compatibility and performance.
I help promote MeowShield. This article is about making that threat-model decision before you touch individual switches.
What a threat model is in this context
You are not protecting "the JavaScript" in the abstract. You are raising the cost of a specific kind of work an attacker can already do because the file ships to a browser or a distributed runtime.
Write down three things before you choose a level:
- What is actually valuable in this file? Distinctive business behavior, a paid feature's client helper, or the internals of an SDK method are typical candidates. A thin wrapper around a public library usually is not.
- What can an attacker already do without reading the source? Call your APIs, replay a public flow, or extract a secret you should never have put in the client. Obfuscation does not close those doors.
- What would a successful copy or tamper actually change for you? A cloned demo is a different problem from a modified entitlement check.
If the honest answer to the last question is "they could charge themselves a different price, skip a paywall, or mint a privileged request," the fix is server-side authorization, not a stronger preset. MeowShield is explicit that no protection is absolute, and that the goal is to increase analysis cost while preserving the behavior you need. Server-side authentication and authorization remain necessary.
What the three presets are for
The public docs describe the presets as baselines in the Playground, not as a ladder you must climb.
Low is for first-time users and compatibility-first projects. The documented defaults keep ES5 Transform and KeepTopVar on, use Normal variable renaming, and enable Unicode plus ConstantSplitting at 50 percent. The intended first step is straightforward: paste a real file, keep those defaults, and generate output.
Medium is described as stronger control-flow protection with moderate cost. The documented starting point is to switch to Medium, keep hp-basic as the control-flow baseline, and leave cfConstantSplitting, cfShuffleState, and cfDynamic off until you have a reason to change them.
High is described as VM-oriented protection for critical business logic. The docs tell you to verify account permissions before you treat it as available. The documented defaults turn instMerge and instShuffle on, leave instClosure and instExtract off, leave VM anti-debug options off, and leave vmAuth off, with vmExpiry prefilled. That last detail matters: a prefilled expiry control is not an invitation to invent a licensing scheme in the client.
If you do not yet know which file is worth protecting, or you have not run the output in the browsers and Node.js versions you support, Low is the honest start. Medium or High can wait until you can describe the extra analysis cost you want and have a way to measure whether the build still behaves.
Choose the file as carefully as the level
A common mismatch is to wrap an entire application bundle at High because a few functions are sensitive. That usually raises download and runtime cost for code that is not what you care about, and it makes failures harder to isolate.
A better first cut is the smallest module that contains the distinctive logic. Leave vendored open-source libraries alone unless you have a specific reason to include them. Keep public exports, documented types, and network contracts stable; those are the integration surface your customers or your own app already depend on.
The official path is also staged: start with a preset in the Playground, understand options in Docs, then move authorized server integrations to the separate API. The API is permission-based, and credentials belong in a server-side or CI environment, never in public browser code.
Verify before you keep the stronger build
After you generate output, do the checks the product itself emphasizes. Run, copy, or download the result and keep a compatibility check visible. Compare the protected file against the original in your own environment: does the behavior you need still hold, and is the runtime cost acceptable on the path a user actually hits?
If the protected code no longer works, official troubleshooting is conservative: lower the protection level or disable some options, then add options back one at a time and re-test. If it still fails, contact support@meowshield.com with the configuration, runtime environment, and error details.
A level you cannot run is not protecting anything. The smallest preset that still meets the threat you actually have, and that still ships, is the one worth keeping.
If you want to try that path with one real file, start at https://meowshield.com.
Top comments (0)