mnemon
par grivn
AI & LLMsRésumé
Persistent memory CLI for LLM agents.
Description
Ce que fait ce skill
Mnemon is a persistent memory CLI for LLM agents. It stores categorized facts, links related memories by semantic or causal edges, and surfaces relevant knowledge on recall. Multiple named stores support separate memory contexts.
Built-in deduplication and graph-style linking lets agents build a queryable knowledge structure rather than a flat list of notes.
Quand l'utiliser
- Retaining user preferences between agent sessions
- Linking a past decision to its downstream consequences
- Querying prior debugging insights when a similar error recurs
- Storing architectural choices so agents don't re-litigate them
- Pruning low-importance memories to keep the store focused
Exemple de workflow
User asks: Remember that we switched from MySQL to PostgreSQL because of JSON query performance
- 1
Run mnemon remember with category 'decision', importance 4, entities 'MySQL,PostgreSQL'
- 2
Check returned semantic_candidates and causal_candidates for related memories
- 3
Evaluate candidates and link the new memory to an existing architecture insight with type causal, weight 0.8
- 4
Confirm stored memory ID and linked edges
Memory stored with ID and causal edge to prior database selection memory
Prérequis
- mnemon binary (install via Homebrew or go install)
Installation
clawhub install grivn/mnemonExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.