Repository navigation
Beyond --readonly: per-agent, per-call authority for the AWS MCP servers #4649
javmann100
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I build mnki, an open-source verification layer for MCP: a stdio proxy (or a hosted gateway) that checks each
tools/callagainst what the calling agent was authorised to do and records the evidence, before the serverruns it. It wraps a server as it is, so it works with the AWS MCP servers without changes.
Your servers are the clearest high-consequence case in the ecosystem. IAM permissions and the
--readonlyflaganswer "what may this credential do"; what I want to understand is the next question your users hit: two agents
share one credential, one may only describe stacks and the other may deploy, and both need a record of which
one did what. Do teams ask you for that, or do they solve it with separate profiles and move on?
I am looking for a 20-minute conversation, not a partnership. Reference implementation:
https://github.com/MNKIAgentOS/agent-trust; integration paths, including observe mode that records what would
have been denied without changing anything: https://mnki.com/docs/integrations. Thank you for the servers; the
centralised IAM story and audit logging are exactly the shape enterprise users need.
All reactions