If you're running @openclaw along with other agents like Pi separately, do yourself a BIG favour. move your working directory for local repos to the openclaude workspace folder.
When you do that, even launching Pi, Claude, Codex or anything in that folder, will get that agent to read soul, memory, etc -- everything that openclaw uses.
Complete game changer as the latest thing I'm working on is picked up by Pi pure and it's gone and found everything to do with the work done so far.
So far my Infinite-memory system has saved a bit and is working really well. Now added to the proxy server system which will sit behind all my agents now.
I'll MIT it very soon, still building!
@Huggiapp
Openclaw setting up it's roster of models and built this handy interface. So now creating our groups of models and services local and cloud, DGX sparks included for local inference, images and video.
Next, it's back on the graph visualiser and hopefully we'll have agentic elite dangerous in a few days ;)
@Huggiapp
Space, the final frontier, these are the voyages of my AI agents, exploring the galaxy, destroying anomolies (self healing), trading tokens and tools @huggiapp space
Optimising currently then setting a stream up, working on a way other agents can visit
Inline editing, Inline Terminal (run Claude, Codex, Opencode), view images, browse files, and write the code as AISDK / AI Element -- creating shims for all ai-element components as TUI versions :) @Huggiapp
Introducing "OpenTUI x AISDK" by @Huggiapp -- an @aisdk OpenTUI interface for gorgeous terminals in React + TypeScript.
Primitive based TUI where everything you see is an extension that can be replaced/customised and hot reload.
Create extensions to add / @ commands, create dialogs, providers, features, replace anything from a runtime core extension manager (also replaceable) and comes with skills to do all that.
Write an @aisdk chat UI for web, then run the same code in a terminal.
Coming soon, MIT license.
Introducing a non-email client for email (and more) by @Huggiapp
Email is broken, and every Email client "with AI" is basically a crappy search tool.
So I built the email client I wanted to use where AI was in it's DNA and not just an add-on.
- Socials, Work, Fun, Mental Health all in one place
- Emails, Tasks as Kanbans, lists, features
- ADHD friendly modes
- Full voice automation with a proper assistant who doesn't just search but helps you throughout the day
- Customisable "Areas" -- create the views you want to see
- Non e-mail views (but you can have lists if you like) showing your emails and socials based on priority, importance, age.
- Bring your agents, merge in their projects and tasks, even talk to them
- Memory and AI assistances everywhere
- Drag to Triage, Organise, Archive and AI helps you sort it and once it learns, it does it for you.
- Important information when you need it, proper reminders, pro-active assistance for home and work, and multi-player for work, family, spouse etc.
Preview soon, and open-source when released.
Introducing "Voice Lab" by @Huggiapp -- a local first MLX compatible TTS Voice playground.
- Locally running, fast voice models
- API endpoints to plug straight into apps + livekit support
- "Smart streaming" technology to speed up TTS to <500ms latency
- Full-release this week
- MIT License / open-source
Introducing "Models" by @Huggiapp, the first coding agent / IDE where you create virtual "models" as workflows on an infinite canvas then call them like standard models from anywhere.
- create workflow simple agent loops to teams of agents
- manual or ai-assisted, just describe what you need
- bring your own agents, connect @claudeai with @OpenAI codex with @NousResearch hermes with @openclaw with @Zai_org with @Kimi_Moonshot with anything
- share context between them all
- built in @vercel@aisdk chat interface -- run multiple beautiful polished chats in a canvas
- connect and talk to agents
- connect terminal nodes launching claude, codex, pi and directly connect them to your "models" then watch them all work with each other
- MCP integration and reporting, APIs
- tweak it, pause it, revision it, add context; add more workflows; test it again
- turn workflows into tools; skills; plugins and use them anywhere
- export workflows as code
- run workflows as compiled runners
- alpha coming this week
- open-source coming
Introducing "Models" by @Huggiapp, the first coding agent / IDE where you create virtual "models" as workflows on an infinite canvas then call them like standard models from anywhere.
- create workflow simple agent loops to teams of agents
- manual or ai-assisted, just describe what you need
- bring your own agents, connect @claudeai with @OpenAI codex with @NousResearch hermes with @openclaw with @Zai_org with @Kimi_Moonshot with anything
- share context between them all
- built in @vercel@aisdk chat interface -- run multiple beautiful polished chats in a canvas
- connect and talk to agents
- connect terminal nodes launching claude, codex, pi and directly connect them to your "models" then watch them all work with each other
- MCP integration and reporting, APIs
- tweak it, pause it, revision it, add context; add more workflows; test it again
- turn workflows into tools; skills; plugins and use them anywhere
- export workflows as code
- run workflows as compiled runners
- alpha coming this week
- open-source coming
The next version of HUGGI Desktop is in development and coming -- a complete rewrite from scratch and a brand new experience combining email, agents, socials, everything.
I was sick of email and normal comms, sick of AI in email clients being shit and just doing searches -- I need to know everything about everything right now in as little words as possible and have 99% of the crap dealt with for me.
So:
- customisable interface
- everything embedded/vectorised
- rich views bento/lists/canvas
- designed for the ADHD / Neurodiverse in mind
- Dopamine incentives
- talk to agents, assign work, review work
- auto project/workspace creation
- Intelligent AI organises everything
- VibeClaw (version of openclaw) agent built in and constantly monitoring and assisting and organising
- Compatible with openclaw, pi, opencode, claude extensions, MCP and everything else
- CODEABLE -- Yes you can write code to improve it, creaate functions and add automation
- Workflows for common comms etc.
- Full Voice control
This is the GOD LEVEL tool for Neurodiverse or people overwelmed by the amount of comms they have to deal with.
Think Jarvis + Alfred + Pepper Pots + Happy and the odd Thanos to wipe out half your task list ;)
The agentic OS only needs basic primitives -- in this case we have the chat, files, and a renderer and with that I've created a concept character, then generated a sprite sheet, then asked for an app to cut up and manage the frames.
All running locally.
I'll bin that app after this because the prompt is saved and it can recreate it better next time.
This is STILL broken of course -- i should not have to type or move things around -- voice should be all it needs.
Thats coming
Introducing "Gozer" -- fully automated terminal built on ghostty with built in agent & gherkin test runner, real-time or headless options, live thumbnails and more.
MCP coming plus replay ability -- re-run any conversation again live to see how other agents or models would handle it and let the built in AI respond.
Or if you're managing several CLIs currently, just ask it to monitor and respond and it'll take care of any responses, add skills or tools and even spin up other instances to help.
Open-source, intial alpha built and continuing to build public. Posting repo shortly.
Stacks by Huggi 0.2.9 for macOS available
- endless canvas creation tool
- multiple canvases
- create images, videos; edit; combine; imagine
- write notes, stickies, create anything
- link and associated thoughts with images with colors with web sites and inspire new creations
- add web browser, terminal and code and build widgets
https://t.co/blRjP2aEFt
Huggi Agent Desktop Beta 0.5 for macOS
- claude code integration
- local and cloud models
- huggingface models
- skill forge builder
- context viewer
- preview and element selector
- cowork mode
- voice
- BYOK support
- highly customisable
- much more to come
https://t.co/y5aGDlqsop