Scheduled agents that remember
A daily digest agent that knows what it already told you.
Without memory
Cron-triggered AI jobs are amnesiacs: the Monday digest repeats Friday’s items, the monitoring agent re-alerts on the anomaly it already flagged, and nothing learns from what you actually clicked, read, or ignored.
With Rekall
Rekall’s scheduler runs agents on a cadence with full memory: each run knows what previous runs reported, what you engaged with, and what changed since. Reports get shorter and sharper instead of longer and noisier.
How it works
Schedule the agent
Cron-style schedules spawn runs of a persistent agent — same identity, same memory, every time.
Each run recalls the last
“Already reported” is a memory query, so updates are deltas: what’s new, what escalated, what resolved.
Engagement feeds back
What you act on teaches the agent what matters; the digest adapts instead of accumulating.
See it in action
[digest-agent] run #61 (daily, 07:00)recall: reported-items last 7 runs → 34 itemsnew since yesterday: 3- CI flake rate up 2% (NEW — not previously seen)- PR #61 merged (you watched this thread)- Qdrant 1.13 released (matches your interests)suppressed: 9 items already reported + unclicked
Related use cases
Agents that improve themselves
Run 50 beats run 1 — reflection turns experience into procedure.
Durable agent runs — checkpoint, resume, audit
A crashed agent resumes from its last checkpoint, not from scratch.
Memory-triggered automations
When memory changes, your systems react — webhooks on decisions, gates, and milestones.