OpenAI: Don't Hide MCP Tools Behind listCommand/executeCommand Dispatchers
bibryam · x · 2026-10-23
From bibryam's keynote takeaways (12/16): OpenAI advises against hiding everything behind listCommand and executeCommand in MCP servers. The extra dispatch step adds latency and makes per-tool access control harder. Instead, expose clear actions like searchcontacts, getcontact and createcontact so the model can discover them directly.
Related event: MCP goes production-scale as OpenAI shares server best practices(12 posts)→
More from coding & agent
- CMU TA built an AI-coded autograder for 1,200 sheets — and became its prompt injection — moarbugs · 2026-10-23
- GitHub Copilot's overlapping Agent mode and Permissions settings confuse users — tristanbob · 2026-10-23
- Garry Tan endorses Aside Browser for agents: a real browser that never gets bot-detected — garrytan · 2026-10-23
- Perplexity ships Decider for Mac, running its 4B decision model locally with no API key — perplexity_ai · 2026-10-23
- Testing Hermes Agent with n8n: Letting Agents Execute Real Automation Workflows — techlatest_net · 2026-10-23
- Only ~56% of MCP Servers Provide Instructions, OpenAI Data Shows — bibryam · 2026-10-23