{"slug": "langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to", "title": "LangChain’s tool routing is a bloated mess. So we built a <10ms local semantic registry to replace it at US Neural.", "summary": "US Neural built Mycelium, a decoupled semantic tool registry that replaces LangChain's AgentExecutor for sub-10ms local tool routing. The system achieves 9.6ms cold discovery latency and 70.7% family-level intent matching using all-MiniLM-L6-v2 and ChromaDB, all running 100% locally.", "body_md": "Look, LangChain was great to get the ecosystem started 0 to 1. But let's be honest: taking LangChain to production with 50+ tools is a nightmare of abstraction hell and brittle if/else logic. The AgentExecutor is slow, and relying on it to route complex tools feels like fighting the framework.\n\nTool discovery shouldn't be an LLM's job, and it shouldn't require a heavy python wrapper. It’s a distributed networking problem.\n\nWhat we built at US Neural:\n\nWe built Mycelium — a decoupled, semantic tool registry. Instead of chaining tools in code, you register them to a local mesh. When a user has a task, the engine routes it semantically (no keyword matching).\n\nThe Benchmarks (No BS):\n\nWe ran this against a synthetic corpus of 100,000 agents to test if local semantic routing is actually viable for production.\n\nLatency: 9.6ms cold discovery (vs 194ms for BM25).\n\nAccuracy: 70.7% family-level intent matching.\n\nInfra: 100% local (all-MiniLM-L6-v2 + ChromaDB).\n\nIf you are tired of bloated frameworks and want deterministic, sub-10ms tool routing, check out our methodology.\n\nGitHub & Reproducible Scripts: [https://github.com/udaysaai/mycelium](https://github.com/udaysaai/mycelium)\n\nVisual Dashboard: [https://mycelium-agents.netlify.app/](https://mycelium-agents.netlify.app/)\n\nI’m ready to be roasted. Tell me why this architecture wouldn't scale for your stack.", "url": "https://wpnews.pro/news/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to", "canonical_source": "https://dev.to/udaysaai/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-registry-to-replace-56f3", "published_at": "2026-07-26 17:05:36+00:00", "updated_at": "2026-07-26 17:30:28.163475+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "machine-learning", "natural-language-processing", "ai-infrastructure"], "entities": ["US Neural", "Mycelium", "LangChain", "ChromaDB", "all-MiniLM-L6-v2"], "alternates": {"html": "https://wpnews.pro/news/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to", "markdown": "https://wpnews.pro/news/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to.md", "text": "https://wpnews.pro/news/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to.txt", "jsonld": "https://wpnews.pro/news/langchains-tool-routing-is-a-bloated-mess-so-we-built-a-10ms-local-semantic-to.jsonld"}}