{"slug": "hermes-memory-providers-a-complete-breakdown-for-new-users", "title": "Hermes Memory Providers: A Complete Breakdown for New Users", "summary": "Hermes has released a comprehensive breakdown of its memory system, detailing eight external providers alongside the always-on built-in memory that requires no setup. The guide provides a direct comparison of providers including Hindsight, Holographic, and Mem0, with published LongMemEval benchmarks showing Hindsight achieving 91.4% retrieval accuracy compared to Mem0's 67.6%. For most new users, the built-in memory handling preferences and project facts is sufficient, while external providers become necessary for advanced use cases like multi-agent setups or production search quality.", "body_md": "Hermes has a *lot* of memory options. If you're new, the choices can be overwhelming — built-in memory, 8 external providers, different costs, different architectures. This guide breaks it all down so you can make the right call for your setup.\n\nBefore we talk providers, understand that **built-in memory is always on**. It doesn't cost anything, requires no setup, and works out of the box.\n\nTwo files in `~/.hermes/memories/`\n\n:\n\n| File | Purpose | Char Limit |\n|---|---|---|\nMEMORY.md |\nAgent's notes — environment facts, project conventions, lessons learned | 2,200 chars (~800 tokens) |\nUSER.md |\nUser profile — your name, preferences, communication style | 1,375 chars (~500 tokens) |\n\nBoth are injected into the system prompt at the start of every session. The agent manages them automatically — it saves preferences you correct, environment facts it discovers, and conventions it learns.\n\n**Key details:**\n\n`§`\n\ndelimiters`MEMORY [67% — 1,474/2,200 chars]`\n\n)**For most new users, built-in memory is enough.** It handles preferences, project facts, and daily workflow notes. You don't *need* an external provider for a personal assistant setup.\n\nBut you'll want one when:\n\nAll external providers are installed via:\n\n```\nhermes memory setup      # interactive picker\nhermes memory status     # check what's active\nhermes memory off        # disable\n```\n\nOr set manually in `~/.hermes/config.yaml`\n\n:\n\n```\nmemory:\n  provider: hindsight    # or any of the 8\n```\n\n**Important:** Only *one* external provider can be active at a time. All of them layer on top of built-in memory — they don't replace it.\n\n| Provider | Storage | Cost | Unique Angle | Best For |\n|---|---|---|---|---|\nHindsight |\nLocal/Cloud | Free (local) | Knowledge graph + reflect synthesis | Highest accuracy, privacy |\nHolographic |\nLocal SQLite | Free | HRR algebra + trust scoring, zero deps | Air-gapped, zero-install |\nOpenViking |\nSelf-hosted | Free (AGPL) | Tiered L0/L1/L2 loading, 80-90% token savings | Self-hosted teams, cost optimization |\nMem0 |\nCloud | Freemium | Server-side LLM extraction, dual memory scope | Fastest setup |\nHoncho |\nCloud/Self | Paid (cloud) / Free (self-hosted) | Dialectic user modeling | Multi-agent, deep user understanding |\nByteRover |\nLocal/Cloud | Freemium | Knowledge tree in human-readable Markdown | Pre-compression knowledge capture |\nRetainDB |\nCloud | Paid | Hybrid search: vector + BM25 + reranking | Production search quality |\nSuperMemory |\nCloud | — | Web-focused memory with browser integration | Web research workflows |\n\nOnly two providers have published LongMemEval scores:\n\n| Provider | Score | Model |\n|---|---|---|\nHindsight |\n91.4% |\nGemini-3 |\nHindsight |\n89.0% | Open-source 120B |\nMem0 |\n67.6% | GPT-4o (LongMemEval-S variant) |\n\nHindsight is the clear retrieval accuracy leader. Others haven't published comparable benchmarks.\n\n**The best all-around choice for most users who want local + accurate.**\n\nStores structured knowledge — discrete facts, named entities, and relationships — not raw text chunks. Its unique `hindsight_reflect`\n\ntool periodically synthesizes higher-level insights across all memories. Think of it as the agent building a personal knowledge graph over time.\n\n```\nSetup:  hermes memory setup → select Hindsight\n        Leave blank for local daemon, or set HINDSIGHT_API_KEY for cloud\nTools:  hindsight_recall, hindsight_retain, hindsight_reflect\nCost:   Free (local PostgreSQL daemon) / Cloud available for teams\n```\n\n**Best if:** You want the highest retrieval accuracy, need structured knowledge, or handle privacy-sensitive data.\n\n**Zero dependencies. Nothing leaves your machine. Literally two tools and done.**\n\nUses Holographic Reduced Representations (HRR) — memories stored as superposed complex-valued vectors. Recall is algebraic, not similarity-based. A trust-scoring mechanism causes confirmed memories to gain weight and contradicted ones to decay over time.\n\n```\nSetup:  hermes memory setup → select Holographic. That's it. No API keys.\nTools:  2 tools (minimal by design)\nCost:   Free. Local SQLite. Period.\n```\n\n**Best if:** You're in an air-gapped environment, hate external dependencies, or want self-correcting memory that learns what's trustable.\n\n**The token-saver. Self-hosted context database from ByteDance.**\n\nIts filesystem-style hierarchy with tiered loading is the standout feature:\n\nThis means **80-90% token cost reduction** vs. loading full context every turn. Auto-extracts memories into 6 categories: profile, preferences, entities, events, cases, patterns.\n\n```\nSetup:  pip install openviking\n        openviking-server\n        hermes memory setup → select OpenViking\n        Set OPENVIKING_ENDPOINT=http://localhost:1933\nTools:  viking_search, viking_read, viking_browse, viking_remember, viking_add_resource\nCost:   Free (AGPL-3.0, self-hosted)\n```\n\n**Best if:** You're running at scale, want self-hosted infrastructure, or need to minimize token costs.\n\n**The \"just make it work\" option. 30 seconds to running.**\n\nServer-side LLM extraction means Mem0's infrastructure decides what to keep. Includes a circuit breaker so memory failures don't block agent responses. Dual memory scope (session + user) means it separates short-term context from long-term facts.\n\n```\nSetup:  hermes memory setup → select Mem0\n        Set MEM0_API_KEY=your-key\nTools:  mem0_add, mem0_search, mem0_get_all\nCost:   Freemium (free tier available)\n```\n\n**Best if:** You want the fastest setup, don't want to self-host, and are okay with cloud storage. Good starting point — you can always migrate later.\n\n**The philosopher. Builds a model of how you think, not just what you know.**\n\nDialectic user modeling captures reasoning patterns, communication style, and decision-making tendencies over time. Two-layer context injection with configurable cadences for refreshes. Supports multi-agent setups with separate AI peers per Hermes profile.\n\n```\nSetup:  hermes memory setup → select Honcho\n        Set HONCHO_API_KEY=your-key\nTools:  honcho_profile, honcho_search, honcho_context, honcho_reasoning, honcho_conclude\nCost:   Paid (cloud) / Free (self-hosted, AGPL-3.0)\n```\n\n⚠️ **Licensing note:** OSS is AGPL v3.0. Self-hosting in a networked app requires releasing your source under AGPL. Using managed cloud avoids this.\n\n**Best if:** You're building a personal assistant that should deepen its model of you over time, or running multi-agent systems with shared user context.\n\n**Your knowledge, stored as readable Markdown. No black boxes.**\n\nHierarchical knowledge tree stored in `.brv/context-tree/`\n\nas human-readable Markdown files. Unique pre-compression extraction hook fires *before* Hermes compresses long conversations, capturing knowledge before context gets summarized away.\n\n```\nSetup:  hermes memory setup → select ByteRover\nTools:  byterover_search, byterover_list, byterover_forget\nCost:   Freemium\n```\n\n**Best if:** You want full visibility into stored memory, or need to capture knowledge from long conversations before compression loses it.\n\n**Search nerd's pick. Hybrid vector + BM25 + reranking.**\n\nCombines multiple retrieval strategies for the highest-quality search results. Vector similarity catches semantic matches, BM25 catches exact keyword matches, and reranking puts the best results on top.\n\n```\nSetup:  hermes memory setup → select RetainDB\nTools:  retaindb_search, retaindb_store\nCost:   Paid\n```\n\n**Best if:** Retrieval quality is your top priority and you're willing to pay for it.\n\n**Web research workflows. Browser-integrated memory.**\n\nDesigned for memory that extends into the browser — captures and retrieves web content as part of your knowledge base.\n\n```\nSetup:  hermes memory setup → select SuperMemory\nCost:   See supermemory.ai pricing\n```\n\n**Best if:** Your workflow involves heavy web research and you want persistent memory of online content.\n\n| Tier | Providers | Notes |\n|---|---|---|\nFree, local |\nHolographic, Hindsight (local), OpenViking | No API keys, no cloud. Holographic is the easiest pick. |\nFree tier / freemium |\nMem0, ByteRover | Start free, pay for higher limits |\nPaid cloud |\nHoncho, RetainDB, SuperMemory | Production features, team support |\nAlways free (built-in) |\nMEMORY.md + USER.md | No setup, always active, 2200 + 1375 char limits |\n\n**Just getting started?**\n\nStick with built-in memory. It covers 80% of use cases. Add an external provider only when you hit its limits.\n\n**Want the best free local experience?**\n\nHindsight (local daemon). Best benchmarks, nothing leaves your machine, structured knowledge graph.\n\n**Want zero config?**\n\nHogrpghic. Pick it in `hermes memory setup`\n\nand you're done. No API keys, no servers.\n\n**Want the easiest cloud setup?**\n\nMem0. 30 seconds, free tier, hands-off extraction.\n\n**Running multi-agent or want deep user modeling?**\n\nHoncho. The dialectic reasoning is genuinely different from every other provider.\n\n**Care about token costs at scale?**\n\nOpenViking's tiered loading will save you 80-90% on tokens.\n\nSwitching is straightforward:\n\n```\nhermes memory setup      # pick new provider\nhermes memory status     # confirm it's active\n```\n\nYour built-in memory (MEMORY.md, USER.md) stays intact regardless of which external provider you use. Note that external providers store data in their own backends — switching providers means starting fresh with the new one's knowledge base. There's no automated migration between providers yet.\n\nDrop them in the comments. I'm happy to help you pick the right setup for your use case.", "url": "https://wpnews.pro/news/hermes-memory-providers-a-complete-breakdown-for-new-users", "canonical_source": "https://dev.to/paper_scratcher_bafb0086c/hermes-memory-providers-a-complete-breakdown-for-new-users-1ki3", "published_at": "2026-05-27 22:40:03+00:00", "updated_at": "2026-05-27 23:22:59.752415+00:00", "lang": "en", "topics": ["ai-tools", "ai-agents", "ai-products", "ai-infrastructure", "large-language-models"], "entities": ["Hermes"], "alternates": {"html": "https://wpnews.pro/news/hermes-memory-providers-a-complete-breakdown-for-new-users", "markdown": "https://wpnews.pro/news/hermes-memory-providers-a-complete-breakdown-for-new-users.md", "text": "https://wpnews.pro/news/hermes-memory-providers-a-complete-breakdown-for-new-users.txt", "jsonld": "https://wpnews.pro/news/hermes-memory-providers-a-complete-breakdown-for-new-users.jsonld"}}