← All tools

Ephemerals Tools

Managed view over your active cached data. Every tool call that produces results stores them in an encrypted, time-limited cache with a cache_ref. Ephemerals tools let you see what's cached, preview it, and understand its shape — without re-running the original tool call.

Try it free

Why this matters

Agents accumulate cache_ref values across a conversation but have no way to ask "what data do I have right now?" or "what does this ref actually contain?" Ephemerals provides that working memory view — agents can list what's cached, inspect the schema and sample rows before deciding what to do, and track how long until entries expire.

When items expire from the cache, they vanish from the list. This is the feature, not a limitation — ephemerals are working memory, not permanent storage.

Capabilities

✓ephemerals.list — List all active cached datasets for the current user. Shows cache_ref, source tool, record count, type, and time until expiry. Sort by newest, expiring soonest, or largest
✓ephemerals.inspect — Detailed stats and preview of a specific cache entry. Returns inferred schema (field names + types), sample rows, record count, expiry info, and source chain (leaf, composite, or view)

Zero cost

Both tools are pure ETS lookups — no LLM, no database, no network call. They read metadata from the same in-memory cache your tool results already live in. No premium credits, no rate limits.

Composes with

Use cache_ref values from ephemerals.list with Data + Frame tools for filtering, sorting, and grouping. Pipe into Prism for charts, exports, or natural language transforms. Use discovery.summary to see the full tool surface before deciding what to do with your cached data.