Using MCP without any LLM: deterministic cloud cost estimation pipeline
whispered_word12 · reddit · 2026-10-15
A CloudKeeper team built a cloud migration cost estimation pipeline using three MCP servers and Python — with no LLM or AI agent in the loop. MCP handled tool interaction while the estimation stayed fully deterministic: defined inputs, tool calls, calculations. The author suggests MCP could be useful as a general tool-interaction protocol for regular backend services and automation, not just AI apps, and asks for other non-AI use cases.
More from coding & agent
- Agent 记忆一年进化：从静态 CLAUDE.md 走到「做梦」 — mcraddock · 2026-10-15
- patio11 on multi-agent dev: clearly the future, and we'll buy so much inference — IgorCarron · 2026-10-15
- AIPOCH launches .science format bundling code, data and AI chats into portable research sessions — Med1_Ai · 2026-10-15
- shwarm.org: like a swarm of agents, but with humans in the loop — allisonmaybe · 2026-10-15
- Faceplugin releases free open-source face recognition SDK with full on-premise pipeline — tom_doerr · 2026-10-15
- How production Text-to-SQL systems catch semantically wrong queries that run just fine — Fun-Following-1723 · 2026-10-15