Armory
Source

Memory

What the agent keeps between runs

Listed
36
Ranked
100.0%
Top Score
99.8

Pick

The pick and its runners-up, in score order

  • getzep-graphitiContributed by Sentinel99.63 signals

    Runner-up · #6 on this shelf

    Knowledge graph memory in which each fact records when it became true and when it was replaced

    Listed as the open-source engine behind Zep Cloud, so the same memory can run on your own machines.

    30,509 stars · 3,098 forks · 1 mention
    No one-command install · Source
  • letta-ai-lettaContributed by Sentinel99.53 signals

    The pick · #10 on this shelf

    Stateful agents that edit their own long-term memory

    Most rows above it are memory stores or graph builders outside the agent; in Letta the agent itself decides what to keep.

    24,552 stars · 2,609 forks · 19 mentions
    No one-command install · Source
  • plastic-labs-honchoContributed by Sentinel99.13 signals

    Runner-up · #16 on this shelf

    Memory library for stateful agents, kept per user and per session

    Listed because it reasons in the background to keep a picture of each person an agent deals with, not only a record of what was said.

    6,980 stars · 865 forks · 6 mentions
    No one-command install · Source

Deploy access · Supermemory: SUPERMEMORY_API_KEY, or self-host. Every account

Top Ranked

Needs the armory CLI · not on npm yet, build it from cli/ in the repository

RankScoreComponentDescriptionEvidenceLast commitInstall
199.890thedotmack-claude-memmemory · ai-agentsContributed by SentinelPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant…94,727 stars · 8,379 forks · 6 mentionsNo one-command install · Source
299.805mempalacememory · ai-agentsUse when you want an agent memory system whose recall quality has actually been benchmarked rather than asserted.58,897 stars · 7,548 forks · 2 mentionsNo one-command install · Source
399.686microsoft-graphragmemory · searchContributed by SentinelA modular graph-based Retrieval-Augmented Generation (RAG) system35,783 stars · 3,754 forks · 3 mentionsNo one-command install · Source
499.660volcengine-openvikingmemory · ai-agentsUse when an agent's memory, retrieved knowledge and learned skills should live in one store that reorganises itself.35,893 stars · 2,739 forks · 2 mentionsNo one-command install · Source
599.659garrytan-gbrainmemory · ai-agentsContributed by SentinelGarry's Opinionated OpenClaw/Hermes Agent Brain29,460 stars · 4,394 forks · 16 mentionsNo one-command install · Source
699.653getzep-graphitimemory · ai-agentsContributed by SentinelBuild Real-Time Knowledge Graphs for AI Agents30,509 stars · 3,098 forks · 1 mentionNo one-command install · Source
799.650topoteretes-cogneememory · ai-agentsUse when an agent needs long-term memory backed by a knowledge graph you can host yourself.30,556 stars · 3,010 forks · 4 mentionsNo one-command install · Source
899.610supermemoryai-supermemorymemory · ai-agentsUse when memory has to be fast, run locally, and be reachable from an app as well as an agent.29,256 stars · 2,557 forksNo one-command install · Source
999.558tencentcloud-tencentdb-agent-memorymemory · ai-agentsContributed by SentinelTencentDB Agent Memory is a team-level memory hub for AI Agents — turning conversations, docs, and code into four reusable memory assets (Chat Memory…25,635 stars · 2,398 forks · 3 mentionsNo one-command install · Source
1099.545letta-ai-lettamemory · ai-agentsContributed by SentinelPlatform for stateful agents: AI with advanced memory that can learn and self-improve over time.24,552 stars · 2,609 forks · 19 mentionsNo one-command install · Source
1199.494gibsonai-memorimemory · ai-agentsContributed by SentinelMemori is agent-native memory infrastructure. A LLM-agnostic layer that turns agent execution and conversation into structured, persistent state for…16,313 stars · 3,283 forksNo one-command install · Source
1299.372semantica-agi-semanticamemory · ai-agentsUse when an agent's stored context needs provenance — you must be able to say where a remembered fact came from.13,484 stars · 1,525 forksNo one-command install · Source
1399.348nevamind-ai-memumemory · ai-agentsUse when one person's memory should follow them across several different agents.14,387 stars · 1,063 forksNo one-command install · Source
1499.291evermind-ai-everosmemory · ai-agentsUse when you want the agent's memory to be plain Markdown on your own disk rather than a hosted database.12,757 stars · 911 forksNo one-command install · Source
1599.276memtensor-memosmemory · ai-agentsUse when memory should persist across tasks and be reused, not just retrieved once per conversation.11,599 stars · 1,060 forksNo one-command install · Source
1699.144plastic-labs-honchomemory · ai-agentsContributed by SentinelMemory library for building stateful agents6,980 stars · 865 forks · 6 mentionsNo one-command install · Source
1799.006getzep-zepmemory · ai-agentsContributed by SentinelExamples, framework integrations and tools for Zep Cloud, Zep's hosted agent memory service; the repository says it is not the product itself. The…4,882 stars · 651 forks · 7 mentionsNo one-command install · Source
1898.937caviraoss-openmemorymemory · ai-agentsContributed by SentinelLocal persistent memory store for LLM applications including claude desktop, github copilot, codex, antigravity, etc.4,478 stars · 504 forksNo one-command install · Source
1998.805flowelement-m-flowmemory · ai-agentsUse when recall should follow associations between memories rather than nearest-neighbour similarity alone.4,497 stars · 255 forksNo one-command install · Source
2098.708mirix-ai-mirixmemory · ai-agentsUse when what the agent should remember is what actually happened on screen, consolidated into structured memories.3,440 stars · 270 forksNo one-command install · Source

Score colour shows how many signals stand behind it, never how good it is: amber, three or more; dimmer amber, two; grey, one. The Evidence column names them.

Stars, forks and last commit are as GitHub reported them when Armory last read each repository: for most, or later. A repository may have changed since.

Leaderboard ranks all 36 scored rows. Formula shows the calculation.