ai-memory
JSON →Persistent memory for any AI assistant. Zero token cost until recall. Stores memories in local SQLite, ranks by 6-factor scoring, returns results 79% smaller than JSON. Works with Claude, ChatGPT, Grok, Cursor, Windsurf, and any MCP client.
Install
pip install ai-memory-mcp Tools · 21
- memory_store Store a new memory (deduplicates by title+namespace, reports contradictions)
- memory_recall Recall memories relevant to a context (fuzzy OR search, ranked by 6 factors)
- memory_search Search memories by exact keyword match (AND semantics)
- memory_list List memories with optional filters (namespace, tier, tags, date range)
- memory_get Get a specific memory by ID with its links
- memory_update Update an existing memory by ID (partial update)
- memory_delete Delete a memory by ID
- memory_promote Promote a memory to long-term (permanent, clears expiry)
- memory_forget Bulk delete by pattern, namespace, or tier
- memory_link Create a typed link between two memories
- memory_get_links Get all links for a memory
- memory_consolidate Merge multiple memories into one long-term summary
- memory_stats Get memory store statistics
- memory_capabilities Report active feature tier, loaded models, and available capabilities
- memory_expand_query Use LLM to expand search query into related terms (smart+ tier)
- memory_auto_tag Use LLM to auto-generate tags for a memory (smart+ tier)
- memory_detect_contradiction Use LLM to check if two memories contradict (smart+ tier)
- memory_archive_list List archived memories (with optional namespace/tier/tag filters)
- memory_archive_restore Restore an archived memory back to the active store
- memory_archive_purge Permanently delete archived memories matching filters
- memory_archive_stats Get archive statistics (counts by tier, namespace, age)
Environment variables
XAI_API_KEY
Links
★ 22 GitHub stars