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

built by @jonkomet who's also hosting an event in Bangkok tomorrow: luma.com/tua74l05

Running Honcho on your own machine? offendingcommit built OpenConcho, a desktop and web UI for self-hosted Honcho. Browse peers, sessions and conclusions, watch dream cycles land, fan one question across every reasoning level side by side, and chat against your own memory. Builds for macOS & Linux, MIT licensed: github.com/offendingcommit/o… Self-hosters keep building the tools they want to use. Post yours in the showcase channel on the Honcho Discord 🫡

we now have over 100 people signed up to come to this event 💫

If you're in Bangkok be sure to check out this event!

This Friday!

Last week we challenged all @plasticlabs employees to devise new ways to visualize what Honcho knows about them. Here are our 4 favorites 🫡 1. Reuniclus from @courtlandleer. Honcho forms a representation of every peer in a workspace. Reuniclus is that whole representation rendered as one living form, every conclusion a point you can explore, interrogate, and replay. 2. Afterimage from @lowyelling. Generative art based on Honcho 's representation of you, one month at a time. Honcho paints you in as the months go by. Replay time, hide layers, take a snapshot. Inspired by the NFT series September by @teaboswell 3. Four Corpora from Phil. 4 agents were asked to write the rules for working with a senior SRE based on 4 different memory systems. Honcho's derived insights held their own against a sprawling agent-wiki, while a working session's context dump pointed towards the data we dream to capture. 4. Honcho Widgets from @akattelu. This experiment pulls a live sample of your Honcho representation & builds a custom dashboard featuring a Vibe Check, where your time goes, conclusions, Spirit Pokémon & more

bringing the full open-weights agent coalition to bangkok this friday. sep 25. doors open 18:00(ICT) at cleverse. 100% free. • desktop app, tui, browser extension, and bot mode live demos • multimodal reasoning & video generation with @MiniMax_AI (m3 + h3) + attendee credits • official memory partner: @honchodotdev deep dive + developer credits • thai voice-first agent showcase with botnoi group • exclusive @NousResearch merch giveaway + compute credit raffle • hands-on onboarding & builder networking over food and drinks hosted by me and officially sponsored by @NousResearch rsvp: luma.com/tua74l05 meetup.com/meetup-group-ltfi…

New in Honcho 3.2 - Now you can check what an answer about a user was built from before you act on it. Pass include_evidence=True on a chat call and the response comes back with every conclusion the agent read, references to the messages it pulled, and the tools it called. This means you can catch a bad answer before it reaches a user, debug a wrong personalization from one request instead of reproducing it, and tell someone exactly what informed a decision about them. It's collated from what the agent actually accessed rather than self-reported by the model, so it's deterministic and costs zero extra model tokens. It does over-report: a conclusion is listed because it was read, which isn't proof the answer leaned on it. Off by default, so nothing changes for calls you already have. Read more in the Changelog - go.honcho.dev/Gr2rMqV

1 week away!

Your agent shouldn’t forget the project when the session ends. Honcho is memory infrastructure for builders creating stateful agents that need to understand people, projects, and conversations over time. It helps you give an agent useful context by storing messages and events, deriving insights in the background, then returning context, search results, peer representations, or natural-language answers. Key features: • Store, reason, query loop – capture messages and events, then retrieve what the agent needs • Prompt-ready context – pull session context and format it for OpenAI or Anthropic message inputs • Python + TypeScript SDKs – add memory through clients for both ecosystems • Coding-agent integrations – connect through MCP or plugins for tools including Claude Code, Codex, OpenCode, OpenClaw, Hermes, and Cursor-compatible clients • Local and self-hosted paths – run a local stack with the CLI or deploy the FastAPI server yourself It’s open-source (GNU Affero General Public License v3.0). Link in the reply 👇

Ive been using @honchodotdev since May I think... its been fantastic for me so far!

Cool, thanks! I'd hate for Hermes to pull a lobster on us. 🤣BTW, Hermes and Honcho are IMO the best combo.

Thanks to everyone who came out to last night's happy hour at the Woods! We had a great time eating pizza and discussing AI, ML & Honcho

Every message Honcho reasons over becomes conclusions about a person. As those accumulate, some of them stop being true. Dreaming is the pass that keeps the picture current. Think of it like sleep for a memory system. The “waking” reasoning process captures what happened. The dreaming process reflects on what it all means. How it works: 👉 Facts that changed get updated. "Works at A" then "works at B" becomes one current fact. 👉 Contradictions get resolved, and the key facts move onto the quick summary Honcho keeps for each person 👉 Patterns get named and assigned a corresponding confidence level Dreams runs on their own once a peer has 50+ new conclusions. On by default, one config flag to turn it off per workspace or session, and you can trigger one from the API whenever you want. Learn more in the docs go.honcho.dev/6GkltAj

180+ signups & counting. Join us in bk tomorrow 6-8pm go.honcho.dev/A6lkHPC