Curtail, Inc Information

Curtail® builds guardrails for AI-generated code. Static tests check your code at rest. ReGrade® checks it in motion. It sends identical real-world requests to both your current and your candidate version, then compares the responses field by field. Tests verify what you expect. ReGrade catches what you don't. The difference is measurable. In a joint study with the Air Force Research Laboratory, 17 AI models from six providers were asked to fix 18 planted bugs in a 300,000-line codebase. Every model fixed more bugs when ReGrade showed it how the software actually behaved, a median improvement of 152% at a median 42% lower cost per bug fixed. The paper and the full data are at curtail.com/research/afrl. It finds what nobody was looking for. One widely-used open-source platform shipped a password hash disclosure bug for seven years while its tests passed every day. ReGrade caught it on the first replay, without being told the vulnerability existed. ReGrade also talks to your agent. Its MCP integration feeds behavioral findings back to your AI coding agent, so the agent corrects itself against what the software really did instead of guessing from a test failure. Underneath is NCAST, or Network Comparison Application Security Testing, our patented method for detecting behavioral differences between software versions. It runs inline as a proxy, handles encrypted traffic, and needs no source code access, CVE feeds, SDKs, or agents. ReGrade catches regressions during development and merge requests. ReCover™ monitors production traffic and diverts it away from failing services when it detects anomalies.