MCP Server - include options for Annotations

Feature Requests
1

My use case:

In our company we are using ambersearch, which offeres to possibility to query MCP servers for company knowledge. Ambersearch always asks for permission to use a MCP tool, if the MCP tool decares itsself as “possible to write”.

N8n only exposes this “write mode” and from my understanding there is no option it set a MCP tool declared as “read only”, or that it auto-detects it, e.g. if there are only “select” statements when e.g. working with SQL tools.

E.g. when using the official n8n MCP node, it only exposes to ambersearch like this, not beeing able to set it to “allow always” as is is identifed as a “write-tool”:

image
image2282×981 175 KB

when using my modification (see below), it exposes like this:

image
image2282×1014 174 KB

Any resources to support this?

The official MCP documentation / specification mentionns ToolAnnotations ; e.g. the “readOnlyHint”. The ToolAnnotations are optional in the MCP schema, but i believe due to my above use case n8n would profit from this feature.

Are you willing to work on this?

I already did!

I cloned to MCP Server Node; modified it to my needs and published it as a community node with the name n8n-nodes-mcp-readonly-trigger .

Be aware that this is primary vibe coded, but works!

Feel free to do some GUI adjustments and to add it to the official node :wink:

2

This is a valid gap - the MCP spec does support readOnlyHint and destructiveHint in tool annotations, and having n8n expose those as configurable options per tool would make workflows much more compatible with permission-aware clients like ambersearch. A simple per-tool config option (read-only / destructive / idempotent) would cover most real-world cases without adding much complexity.