The Claude Code Desktop Companion wraps multiple Claude Code sessions and exposes a plugin system, API proxy routing, and profile-based configuration (example: .claude-desktop-profile.yaml). As a multi-session desktop tool, users running Claude Code agents via your app would benefit from knowing which agents are allowed to operate on their repositories.
AGENTOWNERS provides a deterministic, machine-readable policy file (AGENTOWNERS.yml) that declares which AI agent identities get which permissions (allow / require_approval / block). It's like CODEOWNERS for AI agents — checks in before an agent acts.
Your architecture makes this integration straightforward:
- Plugin System: AGENTOWNERS check could be a pre-tool-use plugin
- Profile Config: A policy file path could slot into
.claude-desktop-profile.yaml alongside existing settings
- Multi-Session: Different sessions (research vs. production) could enforce different policies
For example, a profile could reference:
# .claude-desktop-profile.yaml
profile_name: "Production Safe"
agent_policy: ./AGENTOWNERS.yml
Would you consider adding AGENTOWNERS policy enforcement as a feature or plugin option?
Project: https://github.com/cschanhniem/agent-owners
The Claude Code Desktop Companion wraps multiple Claude Code sessions and exposes a plugin system, API proxy routing, and profile-based configuration (example:
.claude-desktop-profile.yaml). As a multi-session desktop tool, users running Claude Code agents via your app would benefit from knowing which agents are allowed to operate on their repositories.AGENTOWNERS provides a deterministic, machine-readable policy file (AGENTOWNERS.yml) that declares which AI agent identities get which permissions (allow / require_approval / block). It's like CODEOWNERS for AI agents — checks in before an agent acts.
Your architecture makes this integration straightforward:
.claude-desktop-profile.yamlalongside existing settingsFor example, a profile could reference:
Would you consider adding AGENTOWNERS policy enforcement as a feature or plugin option?
Project: https://github.com/cschanhniem/agent-owners