MCP tool chains burn context on intermediate data; letting models write scripts wins on tokens
Future_AGI · reddit · 2026-08-21
A team shares a concrete comparison between wrapping APIs as MCP tools and letting the model write code. On a task that pulled records from one service, filtered them, and forwarded a subset to another, chained MCP tool calls meant a round trip per step, with most tokens spent on intermediate rows passing through the context window. Given the same two APIs and allowed to write a short script, the model finished in roughly one pass, with intermediate data never touching context.
Their rule of thumb: MCP earns its keep for discovery and auth (the parts you want standardized across servers), while for tight multi-step data work, code execution keeps winning on both calls and tokens.
More from coding & agent
- Ex-DeepMind Eng Lead's AI Calling Agent Fo Cut Hang-Up Rate from 80% to 20-30% — julianweisser · 2026-10-16
- SPACE Open-Sources Its Sandbox Platform for Agentic Code Execution — marktenenholtz · 2026-10-16
- Cognition Launches Devin OnCall to Auto-Investigate Production Alerts — DevinAI · 2026-10-16
- Devin Oncall tops SREGym with 100% end-to-end success, auto-resolving production incidents — tianyin_xu · 2026-10-16
- Running a multi-model UI design arena: Paseo orchestrates four design drafts per model — vista8 · 2026-10-16
- Cortex A/B-tests coding-agent skills and rules by replaying developer corrections before keeping them — Samuele Giampieri · 2026-10-16