Native MCP Rollout for Rocky Teams

MCP can make Rocky more useful by exposing specialized tools. Roll it out like production integration work: scoped, tested, documented, and reversible.

Official Hermes Agent docs: https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp

Rollout checklist

Identify the workflow the MCP server improves before adding it.
Prefer least-privilege credentials and avoid sharing broad admin tokens.
Test the server connection with a harmless read-only request first.
Configure only the tools that are needed for the workflow.
Record fallback steps if the MCP server is down.

Operational standards

Name servers clearly so future agents understand what they are calling.
Keep local stdio servers separate from remote HTTP servers in documentation.
Do not publish MCP configuration snippets that contain tokens or host-private details.
After changing MCP configuration, restart or reload according to the docs.
Include MCP behavior in the task’s verification path.

Pitfalls

Adding an MCP server because it is available, not because the workflow needs it.
Giving an unattended job write-capable MCP tools without a narrow scope.
Confusing a successful connection test with end-to-end workflow success.
Letting tool names collide or become ambiguous.

Verification steps

Run a connection or list-tools check.
Exercise one safe tool call in the target workflow.
Confirm error messages are understandable when credentials are missing.
Document exactly which tools are enabled.