Hermes Agent can use terminal tools in the CLI, but not in Telegram. First, check the platform's toolset, not suspect the installation is broken. Different entry points may have different tool permissions, and remote messaging entry points are usually more conservative than local CLIs.
Why CLI and Telegram are different
The CLI allows you to operate directly on your local machine or server, with clearer identities, directories, and permissions; Telegram entry points go through message gateways and may come from mobile phones, group chats, or multiple users. For security, platform toolsets often restrict terminals, file writing, browsers, and high-risk external tools.
If you want Telegram to perform a certain type of task, first confirm the profile, allowed users, and toolset corresponding to this entry. Don't give all your tools to messaging platforms for convenience, especially don't let group chats trigger terminal commands directly.
How to safely release it?
Start with read-only tasks, such as checking status, reading log summaries, and listing to-dos. Only after confirming that only users are allowed to trigger it can gradually open up writing or execution. For actions like deployment, file deletion, or database modification, it is recommended to retain approval or move to the local CLI for execution.
During testing, messages should be sent using real entry points, not just simulated within the CLI. If the CLI works, it only proves the tool itself is usable; Whether Telegram can be called also depends on the gateway configuration, platform permissions, and current session policies. Checking these layers separately makes the inspection much faster.
If you just want to trigger a fixed action on your phone, it's best to encapsulate the action into a low-risk command or script, then let Telegram call this controlled entry point. Don't let the telemessage portal get the full shell directly, as this is difficult to audit and easily misled by context.
