Mnemex
Implements a temporal memory system with Ebbinghaus-style forgetting curves for Claude. Memories decay exponentially unless reinforced through repeated access, with a scoring function that combines use count, time delta, and strength multipliers. Two storage layers: JSONL or SQLite for short-term working memory, and Markdown files with YAML frontmatter for promoted long-term storage. Memories below a forget threshold (default 0.05) get pruned, while those above a promote threshold (0.65) or accessed 5+ times in 14 days move to permanent storage. Tune half-life (default 3 days) and beta parameters to control how aggressively things fade. Good for contexts where you want the assistant to naturally retain frequently-used facts while letting ephemeral details drop away.
- GitHub stars
- 36
- tools
- 14
- transport
- STDIO
- env vars
- 4
- last updated
- Oct 20, 2025
Tools
14 toolsPublic tool metadata for what this MCP can expose to an agent.
save_memorySave new memory with tags, entities (auto-enrichment in v0.6.0+)
Save new memory with tags, entities (auto-enrichment in v0.6.0+)
No parameter schema in public metadata yet.
search_memorySearch with filters and scoring (includes review candidates)
Search with filters and scoring (includes review candidates)
No parameter schema in public metadata yet.
search_unifiedUnified search across STM + LTM
Unified search across STM + LTM
No parameter schema in public metadata yet.
touch_memoryReinforce memory (boost strength)
Reinforce memory (boost strength)
No parameter schema in public metadata yet.
observe_memory_usageRecord memory usage for natural spaced repetition
Record memory usage for natural spaced repetition
No parameter schema in public metadata yet.
analyze_messageDetect memory-worthy content, suggest entities/strength
Detect memory-worthy content, suggest entities/strength
No parameter schema in public metadata yet.
analyze_for_recallDetect recall intent, suggest search queries
Detect recall intent, suggest search queries
No parameter schema in public metadata yet.
gcGarbage collect low-scoring memories
Garbage collect low-scoring memories
No parameter schema in public metadata yet.
promote_memoryMove to long-term storage
Move to long-term storage
No parameter schema in public metadata yet.
cluster_memoriesFind similar memories
Find similar memories
No parameter schema in public metadata yet.
consolidate_memoriesMerge similar memories (algorithmic)
Merge similar memories (algorithmic)
No parameter schema in public metadata yet.
read_graphGet entire knowledge graph
Get entire knowledge graph
No parameter schema in public metadata yet.
open_memoriesRetrieve specific memories
Retrieve specific memories
No parameter schema in public metadata yet.
create_relationLink memories explicitly
Link memories explicitly
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
MNEMEX_STORAGE_PATHdefault: ~/.config/mnemex/jsonlStorage path for short-term memory JSONL files
LTM_VAULT_PATHPath to Obsidian vault for long-term memory
MNEMEX_DECAY_MODELdefault: power_lawDecay model: power_law (default), exponential, or two_component
MNEMEX_PL_HALFLIFE_DAYSdefault: 3.0Half-life in days for power-law decay model
Readme
CortexGraph: Temporal Memory for AI
A Model Context Protocol (MCP) server providing human-like memory dynamics for AI assistants. Memories naturally fade over time unless reinforced through use, mimicking the Ebbinghaus forgetting curve.
Details
- Repository
- github.com/simplemindedbot/mnemex
- Categories
- AI & LLM Tools, Documents & Knowledge, Automation & Workflows
- Package
mnemex- Transport
- STDIO
- Registry
- MCP registry, active
Related AI & LLM Tools MCP servers
View all
Mcp Adr Analysis Servertosin2013/mcp-adr-analysis-serverADR drift detection, content safety, and decision memory for architectural decisions35stars
Lightcmsjonradoff/lightcmsAI-native, open-source CMS (Go) with 128 MCP tools: content, agent sandbox, SEO & AI.34stars
TokenSafetokensafe-dev/tokensafeReduce your AI bills by up to 99% locally: focused code, warm prompt caches & zero secret leaks.33stars
Git Coureralejandro-m-p/git-courerFull git MCP server for LLM agents — 17 tools, AST annotator, auto-backups, real plumbing.32stars
ATXP — Agent Payment & Wallet Infrastructureatxp-dev/atxpAgent wallet, email, phone, and 41 paid tools: search, image/video/music gen, SMS, calls.32stars
DeFi Trading Agentsnirholas/defi-agentsDeFi agent definitions JSON API - crypto trading portfolio management bots32stars