Docs / MCP deep dive

Tools catalogue

Tools exposed by the TokenSaver MCP tools server. Each call is governed by the key's plan and audited in the console.

Chat & models

Tool Purpose
tokensaver_chat Chat completion through the governed pipeline (cache/RAG/compression/PII)
tokensaver_list_models List available provider/model_id ids
tokensaver_get_usage Plan quotas, key metadata, consumption.today/month snapshot
tokensaver_get_consumption Token & USD cost by period (pipeline + egress LLM)
tokensaver_get_quota_status Quotas vs usage, alerts, consumption
tokensaver_estimate_cost Estimate USD cost without an LLM call

Cache

Tool Purpose
tokensaver_cache_stats Redis + semantic cache counts
tokensaver_list_cache_entries List cached LLM responses
tokensaver_cache_feedback Thumbs up/down on a cached response
tokensaver_cache_invalidate Invalidate cache for one prompt
tokensaver_invalidate_cache_run Invalidate cache tied to a pipeline run
tokensaver_cache_purge Destructive bulk cache delete

RAG (admin / inventory)

Tool Purpose
tokensaver_list_rag_documents List RAG documents (requires effective_modules.use_rag / RAG policy)
tokensaver_get_rag_document Get RAG document metadata by id

Pipeline settings & quotas

Tool Purpose
tokensaver_get_pipeline_settings Key pipeline defaults + plan flags
tokensaver_update_pipeline_settings Patch key thresholds (not module on/off — use governance policies)
tokensaver_get_server_pipeline_config Platform-wide pipeline defaults
tokensaver_get_key_quotas Per-key quota overrides
tokensaver_update_key_quotas Update per-key quota caps

Observability

Tool Purpose
tokensaver_get_run Pipeline run details by request_id
tokensaver_list_runs Recent pipeline runs
tokensaver_health API key validity, Redis, embedding health

Chat sessions

Tool Purpose
tokensaver_create_chat Create a named chat session
tokensaver_chat_in_session Multi-turn chat in a session (server-side history)
tokensaver_get_chat_messages List session messages
tokensaver_delete_chat Delete a chat session