MCP server maintainer: agent failures were my tool's fault, not the model's
OrbitWanio · reddit · 2026-10-15
A developer running an MCP server for blockchain data reviewed every major agent failure and found the root cause in his own tool design, not the model:
- Ambiguous options: the tool said a deposit could be paid in one of two coins (10 of one, or 174,276 of another); the agent added them together.
- Instructions lost in delegation: step-by-step notes in the tool's response went to a helper agent, so the main agent never saw them.
- Unclear step boundaries: the tool only fetched a signature; agents assumed it would also claim the reward, which it doesn't.
Each fix was in the tool, not the prompt — the model believed exactly what the tool told it. He asks where to draw the line between model and tool mistakes.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23