Agent with 35 tools got dumber: dynamic tool discovery beat loading everything
whyisthissarah · reddit · 2026-11-19
A team's research agent with 35 available functions kept picking wrong endpoints, passing nonexistent parameters, and failing multi-source tasks. Rewriting tool descriptions didn't help; traces showed tool definitions eating most of the context, with most requests needing only a couple of functions and similar descriptions compounding selection errors. Their fix: expose a few basic commands upfront and let the agent discover specific tools/endpoints on demand (tested with Monid's CLI), caching fetched schemas. Debugging also got easier since traces show which endpoint was discovered and why.
More from coding & agent
- Pi-Adapted Qwen3.8-27B Doubles Local Coding Pass Rate With 63% Fewer Tool Calls — yb2698 · 2026-11-19
- Dirac Robotics Launches CAD-to-URDF, But a Skeptic Says It Costs a Few Euros of Tokens to DIY — kamathsblog · 2026-11-19
- Dev laments copy-pasting between GPT-6.2 Astra and Fable 5.2: 'I'm a meat proxy' — Angaisb_ · 2026-11-19
- Swyx: Build a Business Where New Model Releases Make You Happier, Not Obsolete — js_craft_hq · 2026-11-19
- Dev's ideal multi-model stack: plan with Fable, execute with Sol, review with Astra — jdjohnson · 2026-11-19
- Google launches Gemini 4 Neon: #1 on AA Coding Agent Index, 50-80% cheaper per task — VeryWellVersed · 2026-11-19