Testing 4 Agent Harnesses: Sub-Agent MCP Permission Isolation Fails by Default
PleasantAd9624 · reddit · 2026-07-31
While experimenting with the Model Context Protocol (MCP), a developer uncovered a counter-intuitive security risk: when an orchestrator agent spawns a sub-agent, the child naturally inherits the parent's MCP tool connections rather than being restricted.
Experiment & Findings
- The author tested four major agent harnesses (Claude Code, Codex CLI, Hermes Agent, and Pi) and found that none provided default delegation isolation for MCP tools.
- The only reliable workaround currently available is process-level isolation: spawning separate processes, profiles, or applying explicit tool allowlists for sub-agents. The author calls on the community to discuss best practices for enforcing granular MCP permissions within a single agent session.
More from coding & agent
- AI Agent Autonomously Creates and Updates Math Research Workspace — doodlestein · 2026-07-31
- Dev Reverse-Engineers Giant Traffic Light, Hooks It to HomeKit via Claude Code — aidenybai · 2026-07-31
- ChatGPT autonomously installs Blender and codes a 3D scene render — goodside · 2026-07-31
- Swarms Open-Sources 100+ Multi-Agent Examples Supporting Major LLMs — KyeGomezB · 2026-07-31
- New ComfyUI Node for Krea 2 Fixes Multi-Character LoRA Face Blending — tekprodfx16 · 2026-07-31
- Generating a Complete FPS Game Demo with a Single Claude Prompt — FinanceYF5 · 2026-07-31