MEMANTO Cookbooks — Memory-Powered Application Blueprints
Production-ready blueprints for building AI applications on MEMANTO. Each cookbook covers architecture, setup, and working code for one complete use case.
For day-to-day memory operations (remember, recall, answer), use the memanto skill
instead. Reach for a cookbook when you are standing up a new memory-backed project or wiring
memory into an existing one.
Prerequisites
A free Moorcheh account (console.moorcheh.ai) and the CLI:
pip install memanto
memanto # interactive setup, stores the API key
memanto agent create my-project
Then review:
- Project Setup
- Environment Requirements
Available Cookbooks
1. Persistent Agent Memory
End-to-end setup for an agent with long-term memory across sessions. Agent creation, session lifecycle, the session-start recall pattern, proactive storage, and MEMORY.md sync. Start here — the other cookbooks build on it.
2. Session Continuity
Resume exactly where you left off across restarts. Session token lifecycle and auto-renewal, MEMORY.md as the cold-start snapshot, commitment tracking, and context summarization.
3. Memory Export & Audit
Export, visualize, and audit everything an agent knows. Markdown and OKF bundle export, confidence filtering, tag-based audits, and reversible pruning via expiry policy.
4. Daily Summary Automation
Automated nightly digests. The scheduled job covers daily summary, conflict detection, and the expiry sweep in one pass.
5. Memory-Powered RAG
Question answering grounded in agent memory. The memanto answer pipeline, context window
management, and citation tracking.
6. Conflict Resolution
Detect and resolve contradictory memories before they poison retrieval. LLM-based detection, interactive resolution, and correction provenance.
7. Migrating from Another Provider
Move existing memory into MEMANTO from Mem0, Letta, Supermemory, Langfuse, or an OKF bundle, and verify the import.