
Kizuno18 (@Kizuno18) on X
Thread by Kizuno18 on the move from local stdio/Docker Model Context Protocol servers to remote HTTP/SSE endpoints in enterprise infrastructure. It names three trade-offs: credential boundary inversion (short-lived OAuth/JWT tokens instead of long-lived admin tokens on workstations), tool schema bloat solved by deferred discovery, and governance via HCP Terraform policy-as-code (Sentinel/OPA), audit trails and approvals. It is opinion, with no data or implementation detail.
TL;DR Argues remote MCP endpoints improve credential isolation, tool-schema loading and server-enforced governance compared with local Docker/stdio MCP servers.
Also seen from another angle🧠 AI →
Organisations named here
- blast radius
- policy-as-code
Technologies named here
- Local stdio/Docker-based Model Context Protocol servers
- HCP
- Local Terraform runs driven by an LLM agent
- OPA
- OAuth/JWT bearer tokens
- Model Context Protocol
- Terraform
- Docker
- OAuth
- SENTINEL
Techniques named here
- Credential boundary inversion
- Policy-as-code enforcement at the control plane
- Server-enforced RBAC over authenticated SSE streams
- Role-Based Access Control
Also named here
- Open Policy Agent
Related in AI Security
