Skip to main content

Granular MCP Permissions (Read/Write)

I want dedicated permissions for MCP (Model Context Protocol) that are separate for 'read' and 'write' access. Currently, using MCP requires 'API Access' or 'Manage API' permissions, which might be too broad for certain roles like Product Managers. Having granular MCP permissions would allow us to grant AI tool access without giving full API management capabilities to every role.
Post type
✨ Other
Status: Completed3 comments

Log in to comment and vote

Comments3

  • Markus Palm changed status to Completed
    Team•

    Sep 2

    Pinned

    Hey, quick update - we made a ton of improvements to the MCP, and it now also has granular permissions! :)

  • Jack Finnegan

    •

    Sep 2

    •

    Merged request

    •

    2 votes

    Combined Read/Write MCP Connector for Controlled Content

    I want a combined read + write MCP connector that allows reading from controlled sources (like my own help center and changelog) while also having write access. Currently, the separation makes it impossible to read documentation to write updates in a single pass. Since malicious users can't write to my help center/changelog, this should mitigate the injection risks that the current separation aims to prevent.
  • Alex Z

    •

    Jul 3

    •

    Merged request

    •

    1 vote

    Granular Permissions for MCP server

    Right now, the idea of having a read and write MCP server is slowing down our development. We created a custom MCP server that wraps the APIs and allows us to read and write specific content. This allows our agents to review and update help center articles. It has a very small blast radius and does not allow potentially destructive operations. For example, it can read all of our articles and code, find gaps, and publish drafts for humans to review and publish. Currently, this pattern is difficult to achieve with the Featurebase MCP servers.