How Do Developers Decide Agent Permissions in Prod? Mostly Vibes
sp_archer_007 · reddit · 2026-08-10
The author discusses the pain point of deciding execution permissions when integrating AI agents into real-world business operations like billing.\n\n- Current State: While the industry broadly accepts agents acting in production, almost no one fully trusts an agent to resolve incidents autonomously; humans are still required in the loop for critical steps.\n- Dilemma: There is a lack of standardized processes for determining when an agent can run autonomously versus when human approval is mandatory, with most developers and teams relying on intuition.\n- Discussion: The author calls on the community to share their real-world experiences and engineering practices regarding agent access control and human-agent collaboration.
More from coding & agent
- Stateless MCP Breaks Cross-Call Observability Without Session ID — Thirumalaiboobathi · 2026-08-10
- IWE: An Open-Source Markdown Knowledge Graph for AI Agent Memory — tom_doerr · 2026-08-10
- Dev Tests Autonomous Agent That Ran 3 Weeks Straight, Wrote 800-Page Book — AlexHardy08 · 2026-08-10
- Test: 96 MCP Tools Burn 2,000 Tokens Just for Discovery — Alarming-List-1313 · 2026-08-10
- Raptor: Open-source project turns Claude Code into autonomous security agent — tom_doerr · 2026-08-10
- Hark Handoff Outperforms GPT-5.5 at 90% Lower Cost — Once_ina_Lifetime · 2026-08-10