AI that respects
your boundaries.
Three layers of guardrails ensure agents never modify files they shouldn't, exceed change limits, or violate your coding patterns.
Request AccessThree layers of safety
Global
Platform-wide rules that apply to every agent, every project, every task. Your organization's baseline security posture.
Project
Project-specific overrides that tighten or adjust global rules. Each project can define its own constraints.
Task
Per-task fine-tuning for maximum control. Override any rule for a specific task when needed.
Layers merge with specificity priority: task overrides project, project overrides global.
What you can constrain
File Limits
Set maximum files an agent can change per task. Prevent sprawling, unreviewed modifications.
Line Limits
Cap the number of lines changed. Keep changes focused and reviewable.
Directory Restrictions
Block agents from specific directories entirely. Protect infrastructure, config, and sensitive paths.
Allowed Directories
Restrict agents to specific directories only. Ensure changes stay within scope.
Pattern Enforcement
Detect and reject forbidden code patterns using regex. Catch security anti-patterns automatically.
Naming Conventions
Enforce branch, commit, and merge request naming standards. Consistency without manual review.
Related features
Test your guardrails before deploying
The Play Workbench lets you validate guardrail rules against real changes — before they go live.
Request Early Access