Konvu Guardrails Demo - Enforcing your security guidelines in the AI-Native SDLC

Coding agents changed who writes code. Engineers ship more than ever, and PMs, designers, and support engineers now open pull requests too. Your security guidelines didn't keep up. They live in wikis, threat models, and the heads of your AppSec team, and the agent writing the code has never seen them. ​Konvu Guardrails closes that gap. In this session we'll show it end to end on a real repository: ​Rules drafted from your code and context. Point Guardrails at a repo, add your architecture docs, threat model, or past pentest findings, and it writes down the security invariants your app already depends on, each linked back to its source. ​Your team stays in control. Edit or reject drafted rules, and add your own in plain language, like "every change under billing keeps the approval step." ​- Enforced inside the coding agent. When a change touches a rule, Claude Code, Codex, or Cursor gets the rule and its evidence and fixes the change before anyone reviews it. Whoever is prompting never has to read the guideline. ​- Checked again in CI. Every pull request, including hand-written code, runs against the same rules in report-only or blocking mode, with the decision recorded on the PR. ​- Fixed once, stays fixed. A pentest finding or reproduced exploit becomes a rule, so the same bug can't come back in a new endpoint. ​We'll also cover how Guardrails fits next to what you already run (AGENTS.md files, Semgrep or CodeQL custom rules, AI reviewers, your existing scanners) and leave plenty of time for Q&A.