Hermes Agent has fewer tools after connecting to MCP, which is normal in many cases, not necessarily because the server is broken. Hermes respects the tools.include, tools.exclude, resources, and prompts configurations for each MCP server; If you keep the tool whitelist very narrow, you will naturally only see a small number of tools in the interface.
Investigate by symptoms
| phenomenon | Priority inspection |
|---|---|
| The entire MCP does not load | enabled: false, command path, runtime, HTTP address |
| Only a few tools are missing | tools.include and tools.exclude |
| No resources or prompts are visible | resources: false, prompts: false, or the server itself is not supported |
| After changing the configuration, it is still in the old state | Perform /reload-mcp in a session |
Why is it recommended to narrow down the tool first?
The power of MCP is that it allows Hermes Agent to connect to the file system, GitHub, documentation, internal APIs, and other tools at the same time. But the more tools, the higher the risk of miscalls and permissions. For services with write operations such as GitHub, financial systems, and customer systems, it is recommended to start with only a small number of read-only or low-risk tools, and then gradually open them after confirming that the process is reliable.
The most stable configuration idea for novices
Connect a file system or git MCP first, and only expose one project directory; Then connect to GitHub MCP, and only put low-risk tools such as list_issues and search_code. After confirming that the Hermes Agent can be called correctly, add actions such as creating issues and updating status. Seeing that there are few tools, don't ask if these tools are really needed for the current task.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
