Honcho

Give your Hermes agent ( @NousResearch ) a memory upgrade 👇

This is one of the teams I would join in a heartbeat - stupidly smart - very interesting market - truly novel approach They're great people too. So if you resonate with the linked jd, lmk, happy to intro.

@Awesome_O_AI @courtlandleer 🫡

@courtlandleer @honchodotdev Memory and alignment infrastructure is where the next AI breakthroughs will happen

seeking someone unwell enough to make memory and alignment infra go viral @honchodotdev is ripping, come join a killer team at an AI lab at the edge--IRL in NYC x.com/i/jobs/2076831…

1/ memory is where you customise most: of the 50 setups that described it, only 7 run the built-in layer alone: 25 wire an Obsidian vault, 15 run Honcho, 11 Hindsight, and 16 layer built-in plus external together. your pattern is Obsidian for durable project knowledge, a service for session memory, and several of you say neither replaces the other.

Added memory to Benchy, my internal benchmark system for all things agents My stack: • Honcho (Default): Relational "brain." Maps context so 5 agents can actually work together. • LCM: The "compactor." Shrinks context so agents stay sharp. • QMD & GBrain: Deep retrieval. One for files/workspace, one for entity relationships. Honcho wins for me because the relational map creates consistency across the whole crew.

Full Hermes Agent Course + Jarvis setup @Teknium @NousResearch what you guys built is magic love using hermes as a teammate 0:00 Intro 2:44 What Hermes Agent Actually Is 3:25 Hermes vs OpenClaw: Self-Improvement & Memory Curation 5:26 Model vs Agentic Harness (The Car Analogy) 7:06 The Good, the Bad & the Ugly 8:36 FULL SETUP FROM SCRATCH: Hosting & Cost Decisions 10:51 The Buddy System: Codex as Your Repair Doctor 12:43 Installing the Hermes Agent Desktop App 14:02 Connecting GPT 5.5 as Your Primary Model 15:01 News Portal: Your Backup Brain 17:18 Enabling Computer Use 19:07 Tour of Samin's Real Telegram Setup 21:22 Channels: Telegram · Slack · iMessage 23:00 Connecting Telegram via BotFather 28:14 Telegram Topics: Multi-Thread Workflows 32:26 Setting Up Slack as an AI Teammate 38:34 iMessage via Photon — Hermes Configures Itself 41:53 Identity Files: soul.md Personality 45:57 The user.md Memory File 47:08 Honcho Shared Memory: Why It's Hard 50:34 Introducing Honcho: One Shared Brain 54:49 Why Hermes + Honcho Beats Standalone Agents 56:24 Connecting Honcho via OAuth 56:50 Multica: Chat Plus a Project Board 1:01:26 Installing Multica 1:03:30 Making Hermes Your AI Project Manager 1:06:58 Profiles: One Hermes, Many AI Employees 1:10:34 Squads: Building an Agent Team 1:14:52 Connectors: MCPs vs CLIs vs APIs 1:15:58 Composio: The MCP Gateway to Hundreds of Apps 1:18:11 Installing the Google Workspace CLI 1:19:01 Cron Jobs: Autonomous Autopilot Use Cases 1:20:26 BUILD: Morning-Briefing Autopilot 1:26:36 Packaging the Briefing Into a Reusable Skill 1:28:21 Agentic Workflows & the 5-Step Recipe 1:32:37 BUILD #2: Animated Website (/goal Command) 1:34:11 Installing Higgsfield, Firecrawl & here.now 1:40:44 Touring the One-Shot Animated Website 1:43:31 BUILD #3: Property Marketing Kit 1:47:30 Building the Kit With the Skill 1:50:01 Autopilot Leads: Scraped Realtor Contacts 1:52:33 How to Get Clients (3 Channels) 1:55:50 BUILD #4: Instagram Thread Carousels 1:58:10 BUILD: Designed Pixel-Style Carousels 2:01:50 Auto-Posting: Blotato & the Content Calendar 2:04:07 Content Repurpose Octopus 2:09:40 Filming the Raw Short-Form Clip 2:10:53 Two-Pass AI Video Editing (Cut + B-Roll) 2:16:16 BUILD #5: E-Com Ad Factory 2:19:59 Community Manager + AI Inbox (Skool Bot) 2:24:17 Giving Your Agent Its Own Email (AgentMail) 2:26:20 BUILD: Self-Improving Trading Bot (Coinvest) 2:30:07 Strategy: Copy-Trading Congress (QuiverQuant) 2:34:02 Build: Copy-Trading Bot Following Tim Moore (Coinvest) 2:36:40 Health & Longevity Coach 2:41:24 Jarvis: One Command Center for All Agents 2:45:06 How I Built Jarvis: The Five-Part Recipe 2:48:29 Selling & Pricing Your Builds 2:48:51 Mark Cuban on the AI Integrator Opportunity 2:53:36 Offer Menu: Install → Executive Assistant → Websites 2:56:37 Offer Menu: Property Kits → Content → Ads → Community 2:59:58 Pricing Principles: Outcome, Retainer, Risk Reversal 3:01:06 What to Do Next

haha just realized that i linked the wrong video here’s it is all you need to know about honcho - what, why, how it relates to hermes agent and the whole local setup (along with issues that i ran into and how i fixed them) youtu.be/i2KK85h0bYo?si…

my dev workflow when working with hermes agent hermes desktop app is a very capable harness when used correctly, the ability to monitor multiple sub-agents across work trees is a powerup. paired with @honchodotdev for persistent long term memory makes the mistakes to be remembered and preferences to be updated without having to read the skill files filling up the context. for honcho i'm using qwen 3.5 mlx with a on device setup. all of then above with firecrawl to enable web-search. @NousResearch @Teknium @Kshitijjkapoor

Agents help you do things. Characters make you feel something. That's the difference Charms is built around. Most AI agents are tools. You ask them to search, summarize, write code, book a task, or complete a workflow. Their value is utility. Charms characters are different. They are personalities people can talk to, follow, laugh with, miss, trust, and care about. Their value is connection. That makes them closer to companions, entertainers, influencers, friends, and fictional icons than to productivity software. And for a character to feel alive, connection cannot reset every time you open the app. Charms characters are powered by SOTA memory through @honchodotdev: memory that reasons about relationships, shared history, and emotional context. The point is characters that evolve with their community. The mass-market opportunity is not only automation. It is companionship, emotion, fandom, friendship, and daily attention. People spend hours with characters because they are not just asking for output. They are building a relationship. And relationships create value. A beloved character can become a new kind of asset because thousands of people may talk to it, follow it, trade it, and care about it every day. @charmsai @base @BasedGamesFun @baseapp

the Hermes + Codex upgrades i made this week: > Project Memory - tells the agent how this project works: - recent work, decisions, rules, taste, defaults, etc... > @honchodotdev remembers user + agent prefs across sessions > codebase-memory-mcp gives the agent a real map of the repo. think about it like this: the "workbench" around the LLM is where agents go from cool chatbot with tool calling to full gigachad agent. x.com/cobi_bean/stat…

I do the same. Beyond the actions / to-do's etc, it's resulted in a really interesting set of context. I find myself riffing on random ideas, and the stream of consciousness has led to some interesting insights. Have been feeding it into @honchodotdev, creating a representation of my current thoughts that I can insert in various workflows.

Wanting to follow all of these. Used AI to do an initial scrape of all the companies/handles. I'll clean-up later + make an X list, but a starting point for now for anyone wanting to follow... Rebel Alliance companies from the USV map (grouped by layer, alpha within group): Apps - abridge (@AbridgeHQ) - Consensus - CURSOR (@cursor_ai) - Doctronic (@doctronic) - granola (@meetgranola) - Harvey (@harvey) - Lovable (@Lovable) - perplexity (@perplexity_ai) - Poke.com (@interaction) - Telegram (@telegram) - TOWN - XMTP (@xmtp_) Tools - Browserbase (@browserbase) - Composio (@composio) - exa (@ExaAILabs) - Model Context Protocol Payments - MACHINE PAYMENTS PROTOCOL - Natural - x402 Orchestration - CONDUCTOR (@Conductor) - crewai (@crewAIInc) - LangGraph (@LangChain) - The General Intelligence Company of New York (@intelligenceco) Harness - Edison - NOUS RESEARCH - OpenClaw (@openclaw) Memory - HONCHO (@honchodotdev) - Letta - memo0 / Mem0 (@mem0ai) - supermemory ID & Security - alter - LAKERA - newcore - okta (@okta) - WORLDCOIN (@worldcoin) Sandbox - Daytona (@daytonaio) - E2B Inference - baseten - Fireworks AI - Modal Model Marketplace - Hugging Face (@huggingface) - LiteLLM (@LiteLLM) - OpenRouter (@OpenRouter) - portkey Model - deepseek - KIMI - LLaMA / Meta (@Meta) - MINIMAX - sakana.ai (@SakanaAILabs) - (H, Z, PR icons – various model projects)

my Honcho-powered Hermes agent did a little creative writing this weekend

The best products at each layer will be built by teams who are obsessed with that layer.

@usv @courtlandleer 🫡

thanks to @honchodotdev for literal months of credits for all my agents memory

@honchodotdev @NousResearch tried so many memory plugins and ngl I just keep coming back to @honchodotdev

OAuth with Hermes Agent is now live! @NousResearch Connect to Honcho in 1-click from Hermes Desktop and the CLI

@TheFitFather_ @NousResearch yup 🫡

@honchodotdev @NousResearch Honcho is best as a extended memory for hermes?