Generic API-catalog MCP server vs one tool per job: a context-bloat experiment
MandeepKaur9407 · reddit · 2026-11-16
The author ran a quick experiment: instead of a local Sharp script, they had an agent resize, compress and thumbnail images through a generic API-catalog MCP server (disclosed: ApyHub, their employer). The server exposes only three tools — searchapis, getapispec, callapi — and the agent found the right API for all three jobs on its own.
- The win is tokens: many focused MCP servers keep every tool definition resident in context; the catalog approach stays light no matter how big the catalog grows, since specs are pulled on demand.
- The risk is search ranking: as the catalog grows, you're trusting the right API to surface first.
- The diagram contrasts the two: 3 calls per task with the catalog vs 1 call per task but a tool list that grows with every need.
They ask how others draw the line between one generic server and many small ones.
Related event: MCP Tool Design Debate: General API Catalog Server Tested(2 posts)→
More from coding & agent
- Berkeley paper: LLMs answer with stale values even when the update is in context — ombre-ai · 2026-11-16
- skills-gate: one gate skill fronts 148 agent skills at just 389 tokens — Distinct-Pie2389 · 2026-11-16
- Open-weight Karpo-Decision-31B tops Decision Index at 73.9, one forward pass per answer — CodeByPoonam · 2026-11-16
- shadcn boosts @uisdk: one-shot UI rebuild that's instantly E2E-testable by AI — shadcn · 2026-11-16
- AI agent deletes 43,599 lines of OpenClaw test code in one PR — vincent_koc · 2026-11-16
- JPMorgan shows Jarvis, its internal agent assembly line, at Interrupt NYC — LangChain · 2026-11-16