Skip to main content
Brane is different from prompt guardrails because Brane enforces policy at runtime when an agent attempts an action. Prompt guardrails influence model behavior. Brane controls whether the action is allowed to execute. Both can be useful, but they protect different boundaries.

​
Comparison

​
Why Runtime Policy Matters

If an agent sends an email, deletes a user, issues a refund, queries a database, calls an MCP tool, or writes memory, the consequence happens at the action boundary. Brane evaluates policy at that boundary:
That gives developers a concrete enforcement point.

​
Use Brane With Existing Safety Layers

Brane does not replace:
  • Prompt engineering
  • Model safety systems
  • IAM
  • Database permissions
  • Network controls
  • Secrets management
  • Application authorization
Brane complements those layers by governing agent-specific action patterns.