My use case is that I'd like to use a containerized terminal backend with hermes-agent. The assumption seems to be that the MCP server will run in the host environment, whereas if my agent is running commands in a container, the MCP server there won't have access to the files in the container. I could try to get hermes to run the MCP server in the container and work out how to ensure and configure a consistent URL using a network alias or something, or try to figure out some file-sharing mechanism using volume-mounts, but it's annoying to have to add such extra complexities to my setup just to get codegraph working with it.
It seems like it would be much simpler and more elegant if my agent could just use the excellent CLI directly without MCP.
I can see that you have skills add-lang and agent-eval. Have you considered adding skills for the CLI?
My use case is that I'd like to use a containerized terminal backend with
hermes-agent. The assumption seems to be that the MCP server will run in the host environment, whereas if my agent is running commands in a container, the MCP server there won't have access to the files in the container. I could try to get hermes to run the MCP server in the container and work out how to ensure and configure a consistent URL using a network alias or something, or try to figure out some file-sharing mechanism using volume-mounts, but it's annoying to have to add such extra complexities to my setup just to getcodegraphworking with it.It seems like it would be much simpler and more elegant if my agent could just use the excellent CLI directly without MCP.
I can see that you have skills
add-langandagent-eval. Have you considered adding skills for the CLI?