Agent Policies

This project has been archived. The unified devleaps-policy-client package has been replaced by editor-specific plugins. Use the appropriate plugin for your editor:

• Claude Code: agent-policies-claude-code
• Gemini CLI: agent-policies-gemini-cli

For a server reference implementation, see agent-policies-server.

This repository now serves as documentation for the Agent Policies concept and architecture. The original devleaps-policy-client Python package that shipped as a single client for all editors has been superseded by dedicated per-editor plugins, which offer tighter integration and editor-specific features.

Why Policies

Automating Decisions

Rule files can be forgotten or ignored completely by LLMs. Policies are unavoidable:

decisions[decision] if {
	input.parsed.executable == "terraform"
	input.parsed.subcommand == "plan"
	decision := {"action": "allow"}
}

decisions[decision] if {
	input.parsed.executable == "terraform"
	input.parsed.subcommand == "apply"
	decision := {
		"action": "deny",
		"reason": "`terraform apply` is not allowed. Use `terraform plan` instead",
	}
}

In the image below, the agent is denied by policy from running terraform apply. The agent can then make decisions on what to do next, in this case it outputs that it will attempt a terraform plan as per the policy recommendation.

Automating Guidance

Aside from denying and allowing automatically, policies can also provide guidance through Post-* events:

guidance_activations[check] if {
	endswith(input.file_path, ".py")
	check := "comment_ratio"
}

In the image below, an agent makes a change which includes a comment that is detected as being potentially redundant. The guidance from the snippet is given as feedback to the change, after which the agent by itself decides to remove the redundant comment.