Skip to content
Pacy Devtools

Sending prompts to AI agents

Terminal agents

Section titled “Terminal agents”

These are the class of AI agents that have the most straightforward integration, with the least amount of setup required. Both of these agents run with full access to your system, with the following flags. Please use with caution and understand the security implications.

Claude Code

Section titled “Claude Code”
  • It runs with --dangerously-skip-permissions flag.

Codex

Section titled “Codex”
  • It runs with --ask-for-approval never flag.

IDE agents

Section titled “IDE agents”

Cline, RooCode

Section titled “Cline, RooCode”
  • Since they are fast-moving projects, support for them might occasionally break, but expect it to be restored soon.

Considerations for other AI agents

Section titled “Considerations for other AI agents”
  • Continued support for other AI agents is not guaranteed and is subject to changes in their APIs or policies.

Important note

Section titled “Important note”
  • To receive updates/notifications from the agents, you must enable auto-approval for the Pacy Devtools MCP server. The process varies depending on your editor or extension.