Agent memory systems comparison table
Type: kb/types/note.md · Status: current · Tags: agent-memory
A scannable view of the code-reviewed systems in this collection, generated from
systems.csv. Lightweight (doc-only) reviews are excluded — a
comparison table is for choosing a system, and that calls for code-grounded
evidence. Click any column header to sort (in the rendered HTML site; on GitHub
the raw matrix is itself a sortable, searchable viewer).
This is the human view: one decision per cell. The raw systems.csv
keeps the full one-hot flag sets (every authority mode, every curation operation,
every read-back signal); here each collapses to its single discriminating value,
and a plain-English description leads.
For the findings across the whole population, see the comparison.
How to read the columns
- What it is — a one-line description of the system, lifted from its review. Scan this first; the rest of the row tells you how it works.
- Storage — where memory physically lives: plain
files, a gitrepo,sqlite, anrdbms, avectororgraphstore,kv,in-memory, ormodel-weights. It sets the operational floor — inspectability and diffability at one end, scale and query power at the other. Files-family still leads, but roughly a third are database-backed, and the most common database is plain SQLite, not a vector or graph store. - Read-back — how remembered material reaches the next action: the agent pulls it with an explicit lookup, the system pushes it in unasked, or both. The first question to ask, because it decides whether the agent has to remember to look or whether context arrives on its own.
- Targeting — for systems that push, how the push selects what to inject:
coarse(always-load / session-start, generic recall) versustargeted(the push selects for this instance — an identifier match, or relevance inferred from content by keyword, embedding, or LLM judgment). Pull-only systems push nothing (—). The raw signal one-hots behind this live in the matrix. - Learns from traces — whether memory is mined automatically from the agent's own execution traces rather than authored by hand. It trades throughput for reviewability. Most systems do — and they overwhelmingly push or do both, so automatic learning and automatic activation tend to ship together.
- Enforces — whether the stored memory ever acts as a hard gate (a check the agent can't bypass — a validation that must pass, a blocking rule, a required proof) rather than advisory context it can override. This is the enforcement mode of behavioral authority — nothing to do with authentication. The other authority modes (instruction, routing, validation, ranking, learning) are near-universal and so don't discriminate; they stay in the matrix.
Curation operations and the full authority and signal flag sets are dropped from
this compact view; they live in systems.csv.
The systems (141 code-reviewed)
| System | What it is | Storage | Read-back | Targeting | Learns from traces | Enforces |
|---|---|---|---|---|---|---|
| A-mem | Python memory library with MemoryNote objects, Chroma retrieval, LLM metadata generation, and automatic neighbor evolution | in-memory | pull | — | no | — |
| ACE | trace-derived playbook evolution, reflector-scored bullets, curator additions, optional deduplication, and coarse prompt read-back | files | push | coarse | yes | — |
| Agent Skills for Context Engineering | authored context-engineering skills plus a file-based researcher OS and trace-to-skill example tooling | files | pull | — | yes | — |
| Agent Workflow Memory | web-agent workflow files induced from successful traces and pushed into WebArena/Mind2Web prompts | files | push | targeted | yes | — |
| Agent-R | MCTS trace collection, revision-trajectory synthesis, checkpoint-level read-back, and no runtime retrieval store | model-weights | push | coarse | yes | — |
| Agent-S | GUI agent framework with S1/S2 JSON experience memory, embedding retrieval, S3 reflection, and BBON trace evaluation | files | both | targeted | yes | — |
| AgentFly | planner-executor agent with JSONL case-bank memory, trace-judged case writes, and parametric or SimCSE case read-back | files | push | targeted | yes | — |
| Agentic Harness Engineering | trace-driven outer loop that distills coding-agent rollouts into debugger reports and durable harness edits | files | both | targeted | yes | yes |
| Agentic Local Brain | local PKM capture into Markdown, SQLite, Chroma vectors, mining tables, RAG chat traces, and recommendation ranking | sqlite | both | targeted | yes | — |
| AI-Context-OS | filesystem-first Markdown memory with L0/L1/L2 context loading, generated adapters, MCP/chat read-back, and trace-derived optimization suggestions | files | both | targeted | yes | — |
| ai-memex-cli | Git-backed Markdown vault, agent skill workflows, trace distillation, lint/watch loops, and context bootstrap | files | both | coarse | yes | — |
| ai-modules | deployable multi-vendor skill/plugin bundle with Markdown wiki, session wrapup, task backlog, and linted file memory | files | pull | — | yes | yes |
| Amazon Science SAGE | AppWorld rollouts become reusable Python skills, retrieval state, SFT data, and GRPO reward signal | files | push | targeted | yes | yes |
| Archie | Git-native Arch Linux desktop configuration with Stow deployment, agent briefs, ADRs, and pull-only repo read-back | repo | pull | — | no | yes |
| AriGraph | in-run knowledge-graph world model with episodic observation memory, Contriever retrieval, LLM extraction/refinement, and prompt pushback | in-memory | both | targeted | yes | — |
| Ars Contexta | Claude Code plugin deriving file-based agent knowledge systems with generated context, skills, hooks, trace mining, and coarse push read-back | files | both | coarse | yes | yes |
| Atomic | SQLite-backed markdown atoms with embeddings, semantic graph, wiki/report synthesis, chat, and MCP memory tools | sqlite | both | targeted | no | — |
| Auto-claude-code-research-in-sleep | Markdown skill harness for autonomous research with project research-wiki memory, review traces, and gated trace-derived skill optimization | files | both | coarse | yes | yes |
| auto-harness | benchmark-driven coding-agent loop that mines train traces, evolves agent.py, promotes evals, and gates changes | files | both | targeted | yes | yes |
| Autocontext | iterative evaluation harness with trace-derived playbooks, hints, skills, tools, validators, runtime traces, and optional model distillation | files | both | targeted | yes | yes |
| AutoSci | AutoSci review scoped to its OmegaWiki/SciMem subsystem: file-backed research wiki, schema/runtime contracts, skills, tools, graph, and bounded context packs | files | both | targeted | no | yes |
| Awesome Agent Memory | GitHub README landscape index for agent-memory products, papers, benchmarks, tutorials, articles, and workshops | repo | pull | — | no | — |
| Basic Memory | local-first Markdown knowledge graph with SQLite/Postgres indexes, MCP pull tools, semantic search, schemas, and Claude hook read-back | files | both | targeted | yes | — |
| Beever Atlas | chat-ingestion knowledge base with Weaviate facts, Neo4j graph memory, MongoDB wiki pages, MCP retrieval, and trace-derived wiki synthesis | vector | pull | — | yes | — |
| Binder | local-first typed SQLite graph with Markdown sync, transaction history, CLI/MCP access, LSP validation, and explicit agent reads | sqlite | pull | — | no | yes |
| browzy.ai | terminal personal KB with Markdown/wiki files, SQLite FTS, LLM compilation, query-time context assembly, and trace-derived digests | files | push | targeted | yes | — |
| byterover-cli | local context-tree memory with HTML topic curation, BM25/runtime-signal retrieval, MCP hooks, review logs, dream pruning, and ByteRover cloud sync | files | both | coarse | yes | — |
| cass_memory_system | file-backed procedural memory for coding agents with cass session search, diary summaries, LLM reflection, scored playbook rules, MCP tools, and trauma guards | files | both | targeted | yes | yes |
| Claude Context Guard | Claude Code slash-command memory using project safeguard files, audits, pagination, hooks, and itemised code indexes | files | both | targeted | yes | — |
| Claude Workstream Kit | repo-local active-work memory for Claude Code with ACTIVE.md resume, workstream closure, gates, and verifier agents | files | both | targeted | yes | — |
| claude-obsidian | Obsidian vault memory with agent skills, hot cache, wiki ingestion, hybrid retrieval, locking, hooks, and methodology modes | files | both | coarse | yes | yes |
| ClawVault | deprecated markdown vault memory with graph/search context, OpenClaw prompt hooks, observer compression, facts, and maintenance workers | files | both | targeted | yes | yes |
| Closure-SDK | geometric verification, Closure DNA database state, and experimental carrier-genome memory without agent prompt activation | files | pull | — | no | yes |
| Clude | cognitive memory SDK and MCP server with SQLite/Supabase stores, hybrid recall, dream-cycle synthesis, memory packs, and prompt-file push surfaces | sqlite | both | targeted | yes | yes |
| CocoIndex | incremental AI data-pipeline framework that maintains fresh vector, graph, file, and database indexes for downstream agent retrieval | files | pull | — | no | — |
| Cognee | graph/vector agent memory control plane with session cache, recall routing, trace-derived improve loops, MCP tools, and decorator push | graph | both | targeted | yes | — |
| Compound Engineering Plugin | repo-file workflow memory with generated strategy, brainstorm, plan, solution, pulse, session-history, and review artifacts | repo | both | targeted | yes | yes |
| Context Constitution | authored Letta agent context-management doctrine with documented MemFS, system-prompt learning, progressive disclosure, compaction, and reflection affordances but no local harness implementation | files | pull | — | no | — |
| Continuity | local-first desktop AI workspace with shared SQLite memory, MCP tools, narrative synthesis, prompt push, and org sync | sqlite | both | coarse | yes | — |
| CORAL | filesystem multi-agent coding hub with shared notes, skills, attempts, roles, eval feedback, heartbeat prompts, and worktree isolation | files | both | targeted | yes | yes |
| Cortex | local RDF/SQLite cognitive knowledge service with ontology, hybrid retrieval, MCP tools, reasoning, and access-derived tier learning | graph | pull | — | yes | yes |
| cq | Mozilla AI plugin and MCP store for structured agent knowledge units, review-gated sharing, and agent-led reflection | sqlite | pull | — | yes | yes |
| CrewAI Memory | unified vector memory with LLM extraction, scoped recall, task/HITL learning, tools, and pre-task prompt injection | vector | both | targeted | yes | yes |
| Decapod | Rust repo-native governance kernel with SQLite stores, context capsules, trace lessons, proof gates, and pull-first memory reads | files | pull | — | yes | yes |
| dense-mem | self-hosted MCP memory server with Neo4j evidence, typed claims, verifier gates, fact promotion, and tiered recall | graph | pull | — | yes | yes |
| DocMason | repo-native private-document KB with provenance, governed ask, deterministic retrieval, and interaction-memory promotion | files | both | targeted | yes | yes |
| Dynamic Cheatsheet | trace-derived test-time prompt memory with LLM cheatsheet curation, embedding retrieval, and automatic solver read-back | in-memory | push | targeted | yes | — |
| Echel | project-owned Markdown product memory, deterministic graph/report generation, evidence gates, and task-scoped agent packets | files | both | targeted | no | yes |
| EchoesVault / echoes-vault-opencode | OpenCode plugin that bootstraps a Markdown/Obsidian vault, slash-command read-back, and agent-mediated trace capture | files | both | coarse | yes | — |
| Eidetic | Claude Code Markdown memory with hook-pushed context, FTS/vector recall, trace capture, compounding, drift penalties, and vault export | files | both | targeted | yes | — |
| Engraph | local Obsidian-vault gateway with hybrid search, MCP/HTTP tools, identity context, and folder-feedback learning | files | pull | — | no | yes |
| EQUIPA | SQLite-backed agent orchestrator with trace-derived lessons, episodes, prompt variants, and prompt-time read-back | sqlite | both | targeted | yes | — |
| Exocomp | Go multi-agent coding workbench with YAML roles, .exocomp ledgers, sandboxed tools, skills, and recovery | files | both | coarse | no | yes |
| ExpeL | trace-derived benchmark agent that distills task trajectories into rules and retrieves prior trials as few-shots | files | both | targeted | yes | — |
| Funes | Git-native Librarian protocol for raw-source preservation, compiled Markdown wiki memory, outputs, and health-check governance | repo | pull | — | no | — |
| G-Memory | trace-derived multi-agent memory with Chroma task storage, NetworkX task graph, JSON insights, and orchestrator-pushed examples/rules | vector | push | targeted | yes | — |
| GBrain | Postgres/PGLite-backed agent brain with markdown write-through, hybrid retrieval, graph links, hot facts, skills, and dream-cycle maintenance | rdbms | both | targeted | yes | yes |
| getsentry/skills | Sentry's repo-backed skill and subagent marketplace with authored prompts, routed references, scripts, and validation rules | repo | pull | — | no | yes |
| Gnosis | repo-local why-memory CLI with JSONL entries, disposable SQLite FTS search, doctrine-guided agent capture, and pull-only read-back | files | pull | — | no | — |
| Graphiti | temporal graph memory with episode provenance, LLM extraction, fact invalidation, hybrid retrieval, MCP tools, and pull-only activation | graph | pull | — | yes | — |
| HALO | trace-derived agent-harness optimizer with SQLite desktop trace store, JSONL trace indexes, recursive trace agents, and local analysis runs | sqlite | pull | — | yes | — |
| Hindsight | service-backed agent memory with LLM fact extraction, observations, hybrid recall, integrations, hooks, transfer, and trace-derived learning | rdbms | both | targeted | yes | — |
| HippoRAG | document-ingest memory framework using OpenIE triples, parquet embedding stores, igraph PageRank retrieval, and RAG QA | files | pull | — | no | — |
| hyalo | Rust CLI for structured Markdown vault search, mutation, linting, snapshot indexes, and Claude skill/rule integration | files | pull | — | no | yes |
| HyperAgents | self-improving agent harness where trace-derived benchmark feedback promotes executable patch lineages | files | push | targeted | yes | yes |
| interview-doc-agent | single-file job-document skill using a file-native experience library, templates, and index-guided context | files | pull | — | no | — |
| KBLaM | model-integrated key/value knowledge injection with trained encoders, modified attention, and KB-conditioned generation | files | push | targeted | no | — |
| Kompl | SQLite-backed knowledge compiler that ingests sources into a generated wiki with provenance, FTS/vector retrieval, MCP tools, and chat-derived drafts | sqlite | both | targeted | yes | yes |
| LACP | local control-plane agent harness with trace-derived Obsidian/SMS memory, hook-time context injection, RAG pull, and policy gates | files | both | targeted | yes | yes |
| Letta | stateful agent server with core memory blocks, archival and recall tools, compaction, sleeptime memory agents, and optional git-backed memory | rdbms | both | targeted | yes | yes |
| Link | local Markdown wiki memory with raw captures, reviewed memory pages, bounded query packets, MCP/CLI skills, validation, and local viewer | files | both | targeted | yes | — |
| LLM Wiki (kenhuangus) | local-first Obsidian/wiki compiler with source monitors, LLM extraction and integration, BM25 search, autonomous maintenance loops, and weak trace-derived prompt-optimization scaffolding | repo | both | targeted | yes | yes |
| LLM Wiki (MehmetGoekce) | Claude Code command package for a Logseq/Obsidian wiki with L1 auto-loaded memory, L2 pull queries, schema rules, and lint gates | files | both | coarse | no | yes |
| LLM Wiki (nvk) | portable agent plugin that compiles source files into topic wikis, queryable through index-guided reads, audits, linting, and session lessons | files | both | targeted | yes | yes |
| llm-context-base | markdown-only LLM wiki template with metadata routing, training-period write-back, lint protocols, and multi-agent shims | repo | both | coarse | no | yes |
| llm-project-wiki | prompt-only Claude Code workflow that bootstraps an Obsidian project wiki, wiki-first rules, diff ingest, and gap audits | files | pull | — | yes | yes |
| llm-wiki (Pratiyush) | local file-based transcript-to-wiki compiler with raw session capture, agent-authored wiki pages, static AI exports, MCP tools, and optional synthesis | files | pull | — | yes | yes |
| llm-wiki-coordination | Markdown protocol layer for multi-agent wiki dialogue, consensus, RoleSpace review, and structural audit | files | pull | — | no | yes |
| LLM-WIKI-MCP | local Markdown wiki with SQLite FTS, MCP/CLI retrieval, Ollama ask, provenance ingest, sidecar notes, and ask-history memory | files | both | targeted | yes | yes |
| LLM-Wiki-v3 | Markdown+git wiki with schema validation, provenance-checked ingest, pending review, and hybrid pull retrieval | repo | pull | — | no | yes |
| llmwiki-marimo | local Marimo LLM Wiki with generated Markdown pages, SQLite FTS/citation graph, pull chat tools, and repairable wiki memory | files | pull | — | no | yes |
| Mem0 | memory SDK/server/platform with additive trace extraction, hybrid retrieval, agent plugins, hooks, and pushed context injection | vector | both | targeted | yes | yes |
| Memex | isolated Claude Code runtime that maintains a markdown wiki through queued ingest, query, and lint jobs | files | pull | — | no | yes |
| Memori | SDK and agent integrations with trace-derived augmentation, SQL/Rust storage, hybrid recall, and pre-call memory injection | rdbms | both | targeted | yes | — |
| MemoryOS | hierarchical conversational memory with trace-derived summaries, profiles, knowledge extraction, vector retrieval, and pre-call prompt assembly | files | both | targeted | yes | — |
| MemPalace | local-first ChromaDB/SQLite memory palace with transcript mining, MCP tools, hooks, and explicit wake-up/search read-back | files | pull | — | yes | — |
| memwiki | npm CLI protocol scaffold for AGENTS/hook files, hot-cache wiki memory, agent-maintained trace updates, and coarse push read-back | files | both | coarse | yes | — |
| MentisDB | append-only hash-chained agent memory with MCP/REST tools, skill registry, LLM extraction, and LangChain memory | files | both | targeted | yes | yes |
| Meta-Harness | trace-derived harness search that uses logs, evaluations, proposer skills, and generated code to evolve memory and agent scaffolds | files | both | targeted | yes | yes |
| MiroShark | simulation knowledge graph with Neo4j graph memory, trace-derived agent activity edges, report-agent reasoning traces, and many public export surfaces | graph | both | targeted | yes | — |
| nao | analytics-agent context builder with file-backed project context, SQL guardrails, stories, and pushed trace-derived user memory | files | both | coarse | yes | yes |
| napkin | local Markdown/Obsidian vault CLI and SDK with progressive overview, BM25 search, read, templates, and file writes | files | pull | — | no | — |
| Nuggets | TypeScript HRR fact memory, Pi prompt injection, trace capture, hit-count promotion to MEMORY.md, and Telegram/WhatsApp gateway | files | both | coarse | yes | — |
| o-o | polyglot HTML/bash living documents with embedded update contracts, source caches, changelogs, sync shell, and Claude CLI dispatch | files | pull | — | no | — |
| OKF Harness | local file-first OKF wiki harness with source provenance, bounded CLI reads, generated agent guidance, lint, and graph reports | files | pull | — | no | yes |
| OpenClerk | local JSON runner over Markdown vaults with SQLite projections, provenance, synthesis, reports, and optional semantic modules | files | pull | — | no | yes |
| OpenSage | ADK agent framework with dynamic subagents, Skills, sandbox memory, Neo4j history/memory, plugins, and RL adapters | files | both | targeted | yes | yes |
| OpenViking | context database with viking:// files, session-derived memory, hierarchical retrieval, hooks, MCP, and LangGraph injection | files | both | targeted | yes | yes |
| Operational Ontology Framework | public D+L+A artifact templates for governed AI work, with no implemented memory runtime | repo | pull | — | no | — |
| Origin | local AI-work memory daemon with sourced pages, hybrid retrieval, review gates, git-backed Markdown, and MCP/Claude Code read-back | sqlite | both | targeted | yes | yes |
| OS-Copilot | FRIDAY promotes judged Python execution traces into Chroma-retrieved reusable tools for later planning and codegen | files | push | targeted | yes | yes |
| Pal | Agno AgentOS personal knowledge team with PostgreSQL/pgvector routing, context files, compiled wiki, SQL, schedules, and Agno memory read-back | rdbms | both | coarse | no | yes |
| Phantom | VM co-worker with Qdrant memory, heuristic session extraction, and queued self-evolution over config files | vector | both | targeted | yes | yes |
| pi-self-learning | pi extension that reflects completed agent sessions into git-backed daily, core, and long-term memory files | files | both | coarse | yes | — |
| Playground | TribleSpace pile runtime with user-created temporal memory chunks and budget-aware context cover | graph | both | coarse | no | — |
| Quicky Wiki | document-derived SQLite claim graph with confidence events, metabolism, generated wiki files, and MCP pull tools | sqlite | pull | — | no | — |
| ReasoningBank | trace-derived benchmark memories selected by embeddings and injected into WebArena and mini-SWE-agent prompts | files | both | targeted | yes | — |
| Reflexion | benchmark agents turn failed trajectories and test feedback into task-local verbal lessons for later attempts | in-memory | push | targeted | yes | — |
| REM | episodic memory service with trace-derived episodes, vector/graph retrieval, LangChain injection, and partially wired consolidation | rdbms | both | targeted | yes | — |
| SAGE | consensus-governed local agent memory with MCP turn capture, hooks, hybrid recall, decay, and corroboration | sqlite | both | targeted | yes | yes |
| sage-wiki | LLM-compiled wiki memory with SQLite search/vector/ontology state, MCP pull tools, session capture, and trust gates | files | pull | — | yes | — |
| Secure LLM-Wiki | hardened claim-wiki pipeline with nonce-delimited extraction, trust tiers, adversarial review, write gates, quarantine, and coarse read-back | files | both | coarse | no | yes |
| Self-Training-LLM | offline synthetic Wikipedia QA generation, uncertainty-filtered SFT/DPO datasets, and model-weight learning rather than contextual memory | files | pull | — | yes | — |
| Semiont | event-sourced W3C annotation KB with human/AI peer workflows, graph/vector projections, context gathering, and pull-based agent access | files | pull | — | no | — |
| sift-kg | CLI document-to-knowledge-graph pipeline with LLM extraction, human-reviewed entity resolution, graph queries, and agent skill guidance | files | pull | — | no | — |
| Siftly | local SQLite Twitter/X bookmark knowledge base with AI enrichment, category routing, FTS5 search, LLM reranking, mindmap, exports, and JSON CLI | sqlite | pull | — | no | — |
| Signet AI | local-first daemon memory with SQLite, FTS/vector/graph recall, hook injection, transcript lineage, and guarded repair paths | sqlite | both | targeted | yes | yes |
| SkillNote | self-hosted SKILL.md registry with collections, imports, sync adapters, usage/rating feedback, and prompt-derived draft candidates | rdbms | both | targeted | yes | — |
| SkillWeaver | web-agent trajectories distilled into Playwright API skills with LLM relevance push and verification metadata | files | push | targeted | yes | yes |
| SkillX | trajectory-derived planning, functional, and atomic skill libraries with filtering, merging, and prompt-time retrieval | files | push | targeted | yes | — |
| Smriti-MCP | MCP markdown memory server with file-backed notes, lexical recall, traces, salience, wikilinks, and agent-mediated consolidation | files | pull | — | yes | — |
| Spacebot | Rust team-agent harness with SQLite graph memory, LanceDB hybrid recall, pushed working context, trace-derived persistence, and skill injection | sqlite | both | targeted | yes | — |
| Sparks | Go runtime for Karpathy-style LLM wikis with deterministic ingest, manifest, lint, collection, query, brief, and MCP plumbing | files | pull | — | no | yes |
| Stash | MCP-served Postgres/pgvector memory with episodes, consolidated facts, graph-like relations, goals, failures, hypotheses, and decay | rdbms | pull | — | no | — |
| supermemory | hosted memory API with generated SDK contracts, profile/search injection middleware, MCP tools, browser capture, graph UI, and trace-derived memory | service-object | both | targeted | yes | — |
| Synapptic | trace-derived user-model builder that mines Claude transcripts into weighted profiles, benchmarked guards, and assistant memory files | files | both | targeted | yes | — |
| Synthadoc | local LLM wiki compiler with Markdown pages, provenance, lifecycle audit, query agents, routing, SSE UI, and context packs | files | both | targeted | no | yes |
| Synto | local LLM vault compiler that turns raw notes into reviewed Markdown wiki articles, SQLite identity state, agent packs, and MCP read tools | files | both | targeted | yes | yes |
| Tendril | self-extending desktop agent with filesystem capability registry, model-authored TypeScript tools, and Deno sandbox execution | files | pull | — | no | yes |
| Thalo | structured plain-text knowledge language with schemas, validation, git-aware synthesis actualization, LSP tooling, and PR automation | files | pull | — | no | — |
| TheKnowledge | file-first LLM wiki gateway with citation-grounded Markdown, NotebookLM synthesis, MCP tools, and policy distillation | files | both | targeted | yes | yes |
| Tolaria | files-first Markdown vault desktop app with Git, type conventions, MCP tools, and active-note AI context push | files | both | targeted | no | — |
| Tracecraft | serverless bucket-backed multi-agent coordination with JSON memory, atomic claims, mailboxes, artifacts, and session mirroring | files | pull | — | no | yes |
| Virtual Context | proxy-owned context virtualization with trace-derived compaction, facts, paging tools, and prompt-time memory injection | sqlite | both | targeted | yes | yes |
| VLM-wiki | file-backed multimodal personal wiki scaffold with AGENTS instructions, raw media, VLM analysis scripts, and Obsidian Markdown read-back | files | pull | — | no | — |
| Voiden | offline Git-native API workspace with .void Markdown/request files, linked blocks, request history, extension skills, and local execution | files | pull | — | no | yes |
| Voyager | Minecraft lifelong-learning agent with trace-derived executable skill libraries, Chroma retrieval, curriculum QA cache, and prompt pushback | files | both | targeted | yes | — |
| WeKnora | enterprise RAG and agent platform with document chunks, wiki pages, graph memory, ReAct tools, and push plus pull read-back | rdbms | both | targeted | yes | yes |
| WUPHF | local multi-agent office with git-backed markdown wiki, per-agent notebooks, fact extraction, learning logs, lint, and cited lookup | repo | both | targeted | yes | yes |
| xMemory | trace-derived hierarchical agent memory with JSONL stores, Chroma/BM25 search, semantic themes, graph files, and pull-only retrieval | files | pull | — | yes | — |
| Zikkaron | Claude Code MCP memory with SQLite/FTS/vector storage, predictive write gating, trace-derived consolidation, hooks, and push/pull recall | sqlite | both | targeted | yes | — |