MochiEveryday companion
Personality · Conversation · Long-term memory
NOMIFUN DESKTOP · Your own personal AI companion
Build your Agent. Meet your companion.
From inspiration to action,
bring the power of AI into your everyday life.
The same companion, on your screen and at your side.
Move your pointer and let ideas meetChoose a mode to exploreEvery capability has a place to shine.
A WORLD OF CAPABILITIES
Think. Create. Act.
Bring powerful capabilities together on your PC.
A personality and memories of its own, with a complete Agent's capabilities.
Keep the same companion close, across your desktop, phone, and physical device.
Personality · Conversation · Long-term memory
“You like making complex things a little easier to understand.”
Designed for lasting, "unbounded memory": archiving, compression and retrieval on demand keep the connection going. Repeated work can become reusable Skills.
Meet NomiFun DesktopThis website demo does not create companions or connect devices. Additional desktop windows use more WebView resources. Each chat platform requires its own setup.
PERSONAL AI / BEYOND THE SCREEN
A companion that remembers you, helps you get things done, and keeps the conversation going on your phone. Connect Xiaozhi Yuntai so it can hear you and respond with voice, expressions, and movement in your everyday space.
Existing Xiaozhi Yuntai project demo. The same identity and memory, with a new way to respond.
Personality, memory, and help with everyday tasks.
Connect to your Desktop and keep the conversation going.
Xiaozhi Yuntai adds voice, expressions, and movement.
Physical device integration currently uses Xiaozhi Yuntai. It requires compatible hardware, configured firmware, pairing on a trusted local network, and a running Desktop. Voice, expressions, and movement depend on the board, firmware, and model configuration.
LOOK CLOSER. GO FURTHER.
Your models, knowledge and tools.
A workspace that belongs to you.
Actual workspace component · Preview with test data · Click to explore
01 / AN AGENT, AS YOU IMAGINE IT
From simple chat to a full working partner. Visually combine capabilities, models, and execution policies,
so every Agent can work its own way.
Understand the code, research options, divide the work, and deliver results you can review.
Your context and drafts stay here. Next turn,
switch to the Agent that fits the task.
Help me turn this idea into a working tool.
First, research the use cases and map out sourced options and implementation boundaries.
Messages and drafts stay · Applies next turnAgent switching is available while a regular local conversation is idle. Companion and channel conversations follow their bound Agent settings. This preview describes the current development source; check release notes for available features.
Requirements + AutoWork + IDMM
From one goal to a continuous work loop.
Keep the goal, resources and acceptance criteria in one requirement. You set the direction. Work starts here.
Organize with tags. Keep the next task in sight.
The assigned Agent session claims the next requirement in order. Persisted execution records keep the queue moving.
When the queue is empty, it waits quietly.
Understand the task, use tools and run checks in the original session. IDMM handles recoverable failures and safe decision pauses alongside it.
Work stays in the main Agent session, with traceable results.
Record deliverables, check results and an execution receipt before advancing. Your acceptance criteria stay with the task.
Verifiable delivery makes the next step reliable.
With permission to write requirements, the Agent can submit follow-up work as instructed. Turn results into new goals in a controlled loop.
Continuous work, always within your authorization.
INTELLIGENCE, IN THE RIGHT PLACE.
IDMM handles execution strategy alongside the main task.
Use rules for clear choices. Call for judgment when needed.
Deterministic rules handle recoverable failures and clear, safe options first. Model calls for the main task are still billed by usage.
Idle queues wait. Enabled work loops resume when the backend starts.
Keep the computer and app running, with model services available.
Approvals, errors or configured limits may pause tasks.
Chat, images, video, and audio.
From one conversation to a canvas you can make your own.
Understand goals, use tools, and search knowledge. Bring your work into one Agent workspace.
Help me turn this idea into a complete creative plan.
Website demo only. No messages are sent or models called.
Prompts, references, models, and settings in one place. Generate, edit, and organize as your ideas take shape.

Set up prompts, references, and generation options. Bring images and video onto a timeline to keep creating.

Choose a speech model and voice settings to turn text into audio. Add audio nodes to the same creative canvas.
Arrange images, video, text, audio, and timelines on a canvas you can freely zoom. Build a workspace around your process.
MAKE IT YOURS MINI APPS & HEADLESS PLUGINS
Create in natural language and preview visually. Semi-managed mini apps and headless services let your tools, background capabilities, and automation grow with your needs.
02 / PREVIEW
Interfaces and background services share one extension package. Preview it before deciding how to use it.
This demo does not create or run real plugins.
From chat to desktop, tools to automation. Choose an entry point to see where your extension can contribute.
agent.tool
Give your Agent custom actions. It can choose when to use them and call your extensions in a conversation.
BUILT TO BE OPEN For developers. For every creator.
From a single conversation to an agent team. See how NomiFun connects models, capabilities and real action.
THE RIGHT MODEL FOR THE RIGHT TASK
Configure tasks, models and protocols separately. Connect chat, images, video, audio and retrieval to the right providers, with room to add new protocols.
Explore the sourceChat model
Understand · Reason · CollaborateA design demonstration; no live models or tasks are running.
Local first. Open source. No product telemetry. Models, IM channels and online tools connect to external services according to your settings.
Open code. Come explore it with us.A PERSONAL IDEA. AN OPEN FUTURE.
In October 2025, NomiFun began as one developer's robot brain and coding enhancement system.
From helping machines understand to giving Agents memory, capabilities, and action, the exploration continues.
Solo development begins, connecting coding tools and intelligent conversation to the physical world.
IDMM sidecar decisions, requests and autonomous work loops, remote robots, companion Agents, and knowledge bases arrive in the product.
Composable Agents, a unified plugin core, and a multimodal workspace open up possibilities for users and developers to build together.
ONE ECOSYSTEM. MANY WAYS IN.
Desktop, mobile, robots and cross-network connections. Four open-source projects, each with a purpose.
Composable agents, desktop companions, autonomous work, and multimodal creation—powered by one local runtime.
Meet DesktopDrag left or right to bring the next possibility into view.
1 / 4: NomiFun Desktop
Your computer is the service. Phones and robots connect directly on a trusted LAN, without deploying an additional application server.
For access across networks, self-host Net Infra and use your own transport channels.
A FEW THINGS TO KNOW
The choices are yours.
The boundaries should be clear.
Conversations, settings and memory stay on your PC by default. Models and network tools connect as you configure them.
Go from user to creator. Explore the interfaces and shape the app around the way you work.
Powerful capabilities need explicit permission. You control access to DMs, group chats, tools and devices.
Phones and robots connect directly on a trusted LAN. Use your own optional relay for access across networks.