session-cache

Conversation-scoped caching reuses previously fetched domain data within an active session to avoid duplicate API calls. This applies to all MCP discovery and general data calls, ensuring the same domain yields the same result without re-fetching.

Session caching intercepts MCP discovery calls like get_instructions and list_actions, storing their responses. When the same domain is requested again in the conversation, the cached result is returned instantly, providing clear status cues for data reuse.

Cached conversation data requires a re-fetch when explicit refresh triggers occur or when parameter changes happen. Outside of these specific re-fetch triggers, the session cache enforces per-conversation rules to reuse the previously fetched domain data.

Yes, conversation data reuse reduces latency by avoiding unnecessary re-fetches. By applying per-conversation caching rules to both discovery and general data calls, the session cache returns identical domain results instantly and logs the reuse status for visibility.

Session caching is limited to the active conversation scope. It does not persist across separate sessions, and cached data is bypassed only by explicit refresh commands or parameter changes, meaning stale data persists until a valid re-fetch trigger occurs.