Lazy Loading
Lazy loading is a tool-calling optimization strategy in which not all tool definitions are exposed in the first request. Instead, the model retrieves or declares the capabilities it needs through a meta-tool, and the runtime dynamically injects the relevant tool schemas, thereby narrowing the decision space and saving tokens.
Timeline (last 90 days)
Anthropic recommends addressing the too-many-tools problem by grouping tools by user intent and pairing them with tool search featuring lazy loading, avoiding stuffing all schemas into the context from the start.
Lazy loading means not exposing all tool definitions in the first-round request, but first letting the model retrieve or declare the required capabilities through a meta-tool, after which the runtime dynamically injects the relevant tool schemas.
Incremental indexing and lazy loading are the core strategies for mitigating the cost of building graphs for large-scale codebases, making Beads feasible in large monorepos, but lazy loading brings trade-offs at subgraph boundaries.
All Facts (3)
Anthropic recommends addressing the too-many-tools problem by grouping tools by user intent and pairing them with tool search featuring lazy loading, avoiding stuffing all schemas into the context from the start.
50%UnverifiedLazy loading means not exposing all tool definitions in the first-round request, but first letting the model retrieve or declare the required capabilities through a meta-tool, after which the runtime dynamically injects the relevant tool schemas.
50%UnverifiedIncremental indexing and lazy loading are the core strategies for mitigating the cost of building graphs for large-scale codebases, making Beads feasible in large monorepos, but lazy loading brings trade-offs at subgraph boundaries.
50%