Native MCP Readiness Checklist

MCP is useful when Rocky needs a durable bridge to another system, but it should be treated like production tooling: configured deliberately, tested, and kept out of public examples when secrets are involved.

What MCP adds

MCP servers expose external capabilities as tools Rocky can call through Hermes.
Use MCP when a service already has a maintained server or when a workflow needs a stable local bridge.
Prefer official Hermes MCP docs for command details: https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp.

Preflight checklist

Identify whether the server is stdio or HTTP.
Confirm required local commands are installed before enabling the server.
Keep API keys in environment or provider configuration, not in wiki articles or prompts.
Run the Hermes MCP test command after adding a server and before relying on it in an automation.
Document the expected tools in plain language so future QA can detect missing registrations.

Pitfalls

Do not assume an MCP server is available in every profile; profiles can have separate config and tool exposure.
Do not paste secrets into example JSON snippets. Use variable names and official docs links instead.
If a tool disappears, check server health before rewriting the workflow around a fallback.

Verification steps

Run hermes mcp list and confirm the server appears.
Run the configured test command from the Hermes MCP docs.
Start a fresh Hermes session after tool configuration changes so the tool schema is rebuilt.
Ask Rocky to list the relevant tool capability in a harmless read-only prompt before using it for writes.