How do you secure 15-plus MCP servers without confusing the model?
Dangerous-Tree-6734 · reddit · 2026-07-21
A user running 15+ MCP servers asks how people manage secrets and permissions without turning the setup into a mess.
Their current approach is:
- keep credentials out of tool configs and reference them from a single gitignored file;
- lock down servers that can send, write, or delete anything, and require confirmation before destructive actions;
- leave read-only servers freer to run;
- keep only the servers actively in use loaded, because too many tools confuse the model.
They ask whether people use env vars, a proper secrets manager/vault, client-side or server-side permission scoping, or a gateway/proxy for central auth and logging.
More from coding & agent
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27
- 'A human approves everything' is a trap: approval fatigue and how to actually design safe agents — Efistoffeles · 2026-10-27