Listing Cache

tools/list, prompts/list, and resources/list calls are cached at the gateway with a configurable TTL. The agent gets a merged, deduplicated view of every upstream's capabilities in a single response. Upstream servers are not called on every listing request.

Configuration

stores:
  cache:
    type: memory

bindings:
  north_mcp_proxy:
    type: mcp
    kind: proxy
    options:
      cache:
        store: cache
        ttl: PT5M

ttl is an ISO 8601 duration. The default is PT5M (5 minutes). Set it lower for upstreams whose tool lists change frequently, or higher for stable upstreams to reduce load.

Store Backends

Store	Edition	Use Case
Memory	Community	Single-node deployments

See Stores for full configuration reference.

Keeping tools/list Short

Caching a listing solves how often upstreams get called, not how long the merged list itself gets as toolkits accumulate. See Tool Discovery for keeping tools/list short with eager/cold partitioning, while every tool stays reachable by name or by search.