Pretense vs CodeShield
CodeShield tells you when AI tools have seen your sensitive code. Pretense stops that from happening. Detection after exposure is not protection. Mutation before transmission is.
CodeShield
CodeShield provides visibility into what code your AI tools are sending to cloud providers. It scans API traffic and alerts security teams when sensitive patterns are detected. Detection is step one; prevention requires a different architecture.
Pretense
Pretense intercepts every LLM API call and transforms proprietary identifiers into synthetic tokens before transmission. The LLM never receives real function names, variable names, or class names. Responses are reverse-mutated so developers work with real code throughout.
Feature Comparison
Side-by-side view of how Pretense and CodeShield compare on the capabilities that matter most to security teams.
Why Teams Switch from CodeShield
Detection is not prevention
When CodeShield detects that your AI tool sent a proprietary function name to an LLM, the damage is already done. The name, its context, and your architecture are now in a third-party training corpus. Pretense ensures that moment never occurs. The LLM receives only synthetic identifiers with no semantic connection to your real codebase.
Mutation preserves developer workflow
A detection-only approach forces developers to choose between productivity and compliance. They either ignore alerts or stop using AI tools. Pretense removes that tradeoff. Mutation happens transparently. Developers use Copilot, Cursor, and Claude Code as normal. Real code is protected automatically.
Audit proof vs audit alert
CodeShield produces alerts. Pretense produces proof. The audit log records every mutation: original token hash, synthetic token, timestamp, provider, and request ID. This evidence demonstrates to SOC2 auditors that proprietary code was protected, not just that you were notified it wasn't.
Frequently Asked Questions
Can I run Pretense alongside CodeShield?
Yes. If your organization already has CodeShield for detection and visibility, Pretense adds the prevention layer. Together they provide detect-and-prevent coverage.
What code patterns does Pretense mutate?
Pretense mutates function names, class names, variable names, and method names across TypeScript, JavaScript, Python, Go, and Java. Comments and string literals are preserved verbatim to maintain LLM context quality.
Is the mutation reversible?
Yes. Mutation is deterministic. Every synthetic token maps back to its original identifier. The MutationMap is stored locally and never transmitted. AI responses are reverse-mutated automatically before they reach the developer.
See More Comparisons
Ready to switch from CodeShield?
Install Pretense in 30 seconds. One environment variable. No code changes. Protect every AI tool request from day one.
No credit card required. Free tier available. Local-first, nothing leaves your machine.