Tired of your AI or Hermes Agent forgetting things between chats? Meet EntropicMem! 🌇 I built it with Hermes, for Hermes! Here's 8 reasons to pair it with your Agent Harness: 1: What it is: EntropicMem is a privacy-first memory plugin for Hermes Agent. Local-first. No cloud. No API keys. No accounts. Just your machine, your data, your AI that actually remembers. Built for people who refuse to hand their conversations to Big Tech. 2: Why it's brilliant: • Local-only. Everything stored in SQLite + Markdown on YOUR disk • Zero network calls, so it's air-gapped safe, literally • MIT licensed, free forever, no subscriptions, no upsells • Python stdlib core, no bloated dependency chain to maintain Your secrets stay yours. Full stop. 3: Deep Research Ranking: Perplexity (the benchmark) gave it 8.8/10, ranking 3rd overall for memory systems, behind only Mnemosyne and Chronicle. That's against cloud-hosted competitors with teams of engineers. EntropicMem does it all locally with ~2,000 tests across Python 3.10–3.13. 4: Real features that matter: • Explainable retrieval, so you see why a memory was recalled (relevance scoring, not a black box) • Per-turn prefetch, which automatically loads relevant context before you even ask • Durable facts + episodic summaries + knowledge-graph triples, three layers of memory, all auditable 5: It's a vault, not a black box: • Markdown files you can open in Obsidian/any editor • Interactive graph view for exploring connections • Append-only audit trail, so nothing is secretly deleted • Pre-compression checkpoints + versioning, rollback anytime 6: Built for the security-conscious: • Automatic secret redaction, so API keys, tokens, passwords never leak • PII detection, yes, including 🇿🇦 South African locale support • Prompt-injection warnings that flag suspicious memory retrieval • Gated destructive ops, deletes require explicit confirmation + backups 7: Installation (30 seconds flat): hermes plugins install entropicmem Then add to config.yaml: memory: provider: entropicmem Bootstrap the vault and you're done. Docs: github.com/Ufonik88/Entropic… 8: What makes it different from the cloud options? They store your memories on their servers. You trust their infra. You accept their terms. You accept rate limits. You accept they can read your memories. EntropicMem stores everything on your machine. Always has. Always will. Ready to give your AI a real memory, on your terms? 👉 Install: hermes plugins install entropicmem 👉 GitHub: github.com/Ufonik88/Entropic… 👉 MIT licensed, locally, forever free #EntropicMem #HermesAgent #LocalFirst #PrivacyFirst #AIPersonalization #OpenSource #LocalAI #Privacy #AISafety #Memory #NeuralMemory #Python @NousResearch

🤖 What you doing with Hermes Agent today? For me, it's Front-End day. Finally. I've been building EntropicMem, my memory plugin for Hermes, and for weeks I've been buried in backend wiring. Retrieval logic, storage layers, the unglamorous work that everything else depends on. The problem? The visual graph got left behind and basic. It's the part of the plugin that maps out your memories and projects so you can actually see them connect. Honestly, it's the part that makes memory feel real instead of just some background process running quietly. 🖼️ I kept telling myself I'd get to it once the backend was solid. The backend kept growing. You know how that goes. So today is Sprint 1 for the Front-End. The graph gets optimized! 💪 I'll be posting progress as I go, so if you're curious how a memory graph comes together from nothing, stick around for the ride. And what's everyone else building? I genuinely want to see what people are doing with Hermes Agent. Drop it below, I'll read every reply! 🦡 #HermesAgent #AIAgents #BuildInPublic #AIMemory #FrontEnd

🔥 Kimi's team has been cooking up something special, and I have a feeling they just turned up the heat. If their earlier models are anything to go by, this could be the new king of Long Horizon value. I'm already paying close attention. 👀

Posts like this from @thdxr are the ones I save instead of scroll past. He's walking through how he actually uses Opencode, and he's the one building Opencode. That combination is rare, and it's the reason this is worth the read. You're not getting a feature walkthrough. You're getting the daily setup of the person who designed the tool and runs it against the product itself. Most of us install something, use a fraction of it, and decide the tool is the limitation. The person who built it usually has a very different thing going on underneath. Custom agents handling the repetitive work. Plan mode before anything is allowed to touch the codebase. A few sessions running at once. Saved commands for the prompts he'd otherwise type ten times a day. Two things come out of reading something like this. You see how the product was meant to be used, which is often not how the docs or the landing page frame it. And you pick up habits that aren't tied to this tool at all, they carry straight over into whatever you're building. The tips themselves are solid. The bigger value is watching how someone who lives in this thing all day organizes his work around it. If you build a tool and use it yourself, write this post. People read them. And for everyone running Opencode, what's the piece of your setup you'd refuse to give up?

🦡 A 3,000 line god file isn't refactored in one go. It's refactored one card at a time. Hermes kanban: each change in its own worktree, dependencies set order, a second profile reviews before anything's done. #HermesAgent #Kanban #Refactoring #AIAgents

🚨 700 pages of NSF investigative files just dropped, and the sharpest line is buried in the notes. 👀 Stanford's pre-award office admitted it never independently verified what researchers put in grant proposals. Which would you rather fund? #ResearchIntegrity #NSF #HigherEd 🎓 🔗 scmp.com/news/us/science-tec…

No more silent truncation. EntropicMem now rejects oversized 'remember' calls instead of cutting the tail, raises the content ceiling, makes cold-start plan and digest pointers portable across both boxes, and updates the docs. 🧠🚀 #EntropicMem Try it out and share your feedback 🙏

Looking for an AI chat app that works across your Mac, iPhone and iPad? 📱 This could be the closest thing yet to Dots and Grok Bot, but built for Hermes. It goes well beyond being a simple chatbot. It also works as a full client for your Hermes Gateway, so your conversations and setup stay in one polished app wherever you are. The developer, @_mattwagner_, is very active and genuinely keen on feedback. Community suggestions often make it into new builds, so your ideas could help shape what the app becomes. 🚀 If you want a powerful, polished Hermes experience and a chance to help build something really cool, now is a great time to jump in.

Most multi-agent frameworks are slide deck fantasy 💣 Nobody shows what happens when two agents edit the same file at the same time. Nobody shows the corrupted state, the race conditions, the agent that marks a broken task as done and moves on. Here is what actual local orchestration looks like on my Hermes Agent host. 10 agent profiles on one Linux box, each pinned to a distinct model: DeepSeek, Qwen, Step. Not one model behind 10 names. Different weights, different strengths, different failure patterns. They all work through a single local Kanban board with task dependency gating. Nothing starts until its blockers are merged. Not planned, not in progress. Merged. Every agent gets its own Git worktree. No agent can touch another profile's workspace, branch, or board state. When something goes sideways mid-task, and it will, the damage is one branch that gets deleted. Every task goes through the same loop: ☑️ Claimed on the board ☑️ Executed in an isolated branch ☑️ Verified against real execution output, meaning tests actually ran and passed, not the agent saying it looks good ☑️ Merged, and the board state updates itself That last part is the filter. If your agents are not writing code, running tests, and updating their own board state without you in the loop, you do not have agents. You have a chat window with extra steps and a bigger bill. The hard part was never the prompting. It was the verification layer and the isolation. An LLM with a terminal and no constraints is not orchestration, it is a liability with API access. What does your local agent stack look like right now? Genuinely curious what people are running past the demo stage 🤔 #AIAgents #HermesAgent #MultiAgent #AgenticAI #DevTools

Verification passes are important, kids! 💡 Sunday maintenance ran overnight and my agent decided to verify the backups. All of them. Weekly archive: 4.76 GiB tar, 155,386 members walked one at a time, gzip -t clean, GPG AES-256/SHA512, then a full decrypt stream to /dev/null. Passed. Then it found two things I did not know about. The desktop daily tar from Oct 11 stopped at 1.6 GB. The daily script backgrounded itself and cron reaped it mid-write. Re-ran it: 5.1 GB, 155,421 files, verified. The other one is worse. GDrive uploads have been dead since Oct 7. rclone.conf is just gone after the host rebuild on the 6th, mygdrive remote sitting there unauthenticated. Five days, no alert, nothing. Both came out of one verify pass. A backup you never verified is just a file you are hoping works.

Hey @NousResearch , I heard you guys had been busy in the lab. I believe someone is missing, no? 😏

Let the games begin! What’s your guess for what Missingno is?

Been using it with my Hermes Agent, running general tasks and research, for about an hour or two. I must say, I’m really enjoying it! I’ve always found @StepFun_ai to have really nice agent models! 💪

Oh??? Is my Hermes bugging out or is that a thing @NousResearch @Teknium ?

I can vouch for Vory. I have been testing it with Hermes and using it for the past few days, and I have been really impressed. 🤯 Voice conversations are especially seamless, and @_mattwagner_ is always happy to hear suggestions and provide quick fixes. Honestly, take 5 minutes to set it up and check it out. 🚀 #Vory #VoiceAI #TechTools #HermesAgent