← All products
🧠

Agent Memory System Kit

Your AI agent forgets everything between sessions, then gets overconfident when memory gets stale. This kit gives it a 3-tier persistent memory system with automatic compression, semantic search, skeptical-memory rules, and CLAUDE.md integration.

✓Updated with the memory hygiene rules I use now.
Updated May 11, 2026 · Added layered memory hygiene, skeptical-memory rules, and anti-bloat maintenance notes.

Paid Digital Thoughts subscriber?

Yearly subscribers: all products free ($1150.94+ value)

Monthly subscribers: 1 free product per month — subscribe from $29.99/mo

Claim your free copy →
$39One-time purchase.

What changed in this update

  • ✓Added layered-memory-hygiene.md
  • ✓Added skeptical-memory-rules.md
  • ✓Added CHANGELOG.md and updated quickstart guardrails

Best for

  • +Agents that need project context across sessions without huge prompt files
  • +Builders who want memory to stay useful after weeks of work

Not for

  • -Teams looking for a hosted memory database or vector platform

What you get

+3-tier memory architecture (daily → weekly → permanent)
+Automatic rollover — old entries compress, recent ones stay detailed
+Semantic search — find relevant context by meaning, not just keywords
+Layered memory hygiene — route facts to the right memory layer instead of saving everything
+Skeptical memory rules — remembered facts must be verified before customer-facing or operational changes
+CLAUDE.md integration snippet — agent reads and writes memory automatically
+setup.sh — one-command installation
+Works with Ollama (free, local) or OpenRouter (remote)
+LaunchD/cron templates for automated daily rollover

Package includes

  • •setup.sh — creates full memory directory structure
  • •rollover.py — automated compression (3-day, 10-day cutoffs)
  • •memory-search.py — semantic search with local or remote embeddings
  • •Memory templates (memory.md, memory-weekly.md, memory-index.md)
  • •layered-memory-hygiene.md and skeptical-memory-rules.md
  • •CLAUDE.md snippet for memory behavior
  • •LaunchD plist template for macOS scheduling
  • •Populated examples showing real memory in action
  • •Comprehensive guide (9 sections, architecture to production)
  • •CHANGELOG.md

FAQ

Does this require an API key?

Core functionality (templates, rollover) works with no API key. Semantic search works with Ollama (free, local) or OpenRouter (paid, remote).

How is this different from the memory in the Blueprint?

The Blueprint includes memory templates as part of a broader agent setup. This kit goes deep: standalone search, automated compression, indexing, and production patterns from 3+ months of use.

Will this work with agents other than Claude Code?

Yes. The memory files are plain markdown. Any LLM that reads instruction files can use them. The rollover and search scripts are standalone Python.

Related reading on Digital Thoughts:

Secure checkout by Stripe. Instant download + guided Claude Code setup.