Digital Life,
Built to Evolve.

AI companions. Digital life agents.
Bringing AI + MR into the world we share.

Our vision / AI + MR

Bringing virtual
creatures to life.

AI agents understand and act. Spatial computing creates a shared space for virtual creatures to interact with the physical world.

A new life, in the world we share.

Bringing virtual creatures to life through interaction with the physical world. AI provides understanding and action; MR connects them to people, objects and places.

Our AI + MR approach

AI + MR fusion is our foundation: AI supports understanding, memory and action, while mixed reality connects an agent to bodies, places and people. AI companionship is a core application alongside games, spatial interaction and everyday services. Lifelike character design connects intelligence with an expressive, recognizable presence that people can help shape.

AI gives behavior context.

Perception, memory and intent help an agent understand and respond.

MR gives presence a place.

Mixed reality brings bodies, gestures and shared physical space into the interaction.

Together, a continuous loop.

People, agents and environments shape what happens next through action and feedback.

AI + MR companionship

Beyond conversation.
An AI Companion across everyday life.

AI brings understanding and continuity. MR gives your companion a presence in the world you share.

A person relaxing at home with a phoneAI + MR concept interaction
AI Companion

“How did you sleep?”

An interactive future scenario for AI + MR companionship. Device signals and integrations are illustrative.

Original connected-life concept

Our worlds

A little closer to life.

AI companions to meet. MR worlds to step into.
We combine AI with mixed reality to connect intelligent behavior, expressive characters and shared physical space.

The colorful cast of Umbrella Up gathered in a playful world

Game / Umbrella Up

Big personalities.
A world of play.

A playful side of Monstar, brought to life through a colorful cast of characters and shared interactive experiences.

Visit Umbrella Up
About Umbrella Up

Umbrella Up is one of Monstar’s published games. It brings character design and embodied play into the same world, extending our exploration of how people connect through interactive experiences.

Glowing Fensy jellyfish float through a dreamlike room with an open skylight

Game / Fensy

A small touch.
A deeper connection.

Meet a luminous companion in dreamlike spaces. Fensy explores MR hand-driven interaction, responsive bodies and spatial storytelling, informing the embodied side of AI + MR digital life.

Explore Fensy
About Fensy

Fensy is one of Monstar’s published games. Its soft-body jellyfish character connects moments of contact, exploration and shared travel. The project also informs our research into embodied interaction.

Mochi: perception, memory and embodied response

Mochi is an early exploration of MR digital life. A proposed interaction loop connects voice, language, gaze, gesture and scene context to intent, interaction memory and behavior selection, then to voice, expression, movement and spatial placement.

  1. SenseVoice, language, gaze, gesture and scene context
  2. InterpretIntent, interaction memory and behavior selection
  3. RespondVoice, expression, movement and spatial placement

User response and interaction history feed the next encounter.

Original framework
Fensy: hand-driven soft-body interaction

Tracked hand-joint displacement and acceleration provide an equivalent-force control proxy. Motion maps to control targets and positional constraints; a vertex-position solver combines structural and shape-matching constraints with volume preservation.

The spatial narrative follows contact, rescue, exploration, shared travel and return, across a dream room, inverted city, desert, grassland and galaxy.

  1. ObserveTracked hand-joint motion
  2. EstimateEquivalent-force control proxy
  3. ConstrainStructural, shape and volume targets
  4. CorrectVertex-position solve

Implicit-force control and XPBD inform the proposed hand–jellyfish coupling. Framework references: XPBD (2016), shape matching (2005).

Original framework

Step into our worlds.

Monstar product film / 1:20

A look at our characters, interactive experiences and future explorations.

Monstar Engine

More than a character.
A sense of presence.

Monstar is building an AI + MR platform for AI digital life agents. AI supports perception, memory and intent; MR gives those behaviors a body and a place in the world we share.

A little curiosity.

Move your pointer. Tap to say hello.

Perceive the moment.

AI interprets voice and vision alongside MR gestures and spatial context, forming a multimodal view of the moment.

Remember what matters.

Identity, interaction history and user preferences provide a basis for continuity over time.

Turn intent into action.

Tools, executable skills and device interfaces connect decisions with actions in a shared environment.

Give behavior a body.

MR connects voice, expression, movement and attention to a body in shared space; AI gives that body context and intent.

Devices

MR / XR headset: A shared spatial interface for gaze, gesture and embodied interaction.

Products &
scenarios

Learning / Virtual characters / Workplace / Wellbeing / Culture & travel / Smart home / Business

Monstar
Engine

Multimodal perceptionVision and sound
Emotional connectionExpression and resonance
User understandingMemory and decisions

2D embodiment

Images, video, voice and expression

Mochi embodiment

3D embodiment

Models, motion and MR presence

A 3D digital-life character

AI infrastructure

Language modelsComputeCloud platformsNetworksData

AI + MR platform direction. Device connections and application scenarios vary by project and stage.

Original framework
Explore the four platform layers

Devices

Companion robots, XR headsets, computers and tablets, phones, smartwatches and smart speakers.

Products and scenarios

Mochi, Umbrella Up and Fensy sit alongside possible applications in learning, virtual characters, the workplace, wellbeing, culture and travel, smart homes and business services.

Monstar Engine

Multimodal perception, emotional expression and user understanding connect to 2D embodiment through images and video, and 3D embodiment through models and motion.

AI infrastructure

Language models, compute, cloud platforms, networks and connected data support the layers above.

From reusable assets to runnable agents

Our planning explores digital life as a versioned project package: model configuration, persona, memory, prompts, tools, device interfaces and evaluation definitions belong together.

  1. ManifestRead configuration, dependencies and resources.
  2. DependenciesResolve model, tool and runtime requirements.
  3. SandboxRun the experience in an isolated environment.
  4. EvaluateCheck behavior, interaction and continuity.
  5. Fork and deployAdapt a version and share it with a community.

Architecture concept from the planning materials; this is not a claim that every runtime capability is publicly available.

Platform roadmap

Community knowledge

Collect interaction ideas, examples and tutorials that people can discover, save and reproduce.

Reusable runtime capabilities

Develop standard project structures, configuration, version management and capability orchestration, with reproducibility and compatibility as evaluation goals.

Personal intelligence

Explore long-term memory, preference learning and coordination across devices, so an agent can become more relevant through continued interaction.

Long-term direction from the September 2026 planning materials; stages describe intentions, not a release schedule.

Research direction

What does it mean
for digital life to evolve?

How can AI digital life agents gain new capabilities while remaining recognizable and open to human direction? Our research direction examines this through AI + MR encounters, shared spaces and human-guided change.

Can new executable skills persist and transfer?

Propose a candidate executable skill.

Evaluate: Transfer, retention and cost

One shared research loop

  1. ObserveInteraction
  2. GenerateCandidate skill
  3. TestNew and existing skills
  4. ChooseHuman direction
  5. RetainVersioned change
  6. ReuseNew context

Memory / Executable skills / Body interfaces / Version history

Proposed research agenda. The mechanisms and studies remain to be evaluated.

Original framework
Situated autonomy: the research framework

Situated autonomy is a working research frame for how digital life perceives context, forms intentions, acts in the world and maintains a sense of being the same agent over time.

Its central proposition connects contextual input with consistent, embodied and memory-bearing behavior. The three research axes are behavior generation, long-term consistency, and scale and ecosystem. Mochi, Fensy and Umbrella Up provide different project probes within this broader program.

A research framework and design proposition, not a reported experimental finding.

People and the ecosystem

People shape
their digital life.

The desire for connection can turn people into builders, combining identity, memory, models, sensing and tools around an agent of their own.

Character & identity

Character cards, worldbooks, prompts, roles and expressive traits.

  1. CharacterIdentity and traits
  2. WorldbookPlaces, events, people and rules
  3. PromptRole, tone and style
Memory & continuity

External memory, context migration and synchronization help preserve a recognizable AI companion.

  1. StoreConversations, images and preferences
  2. RetrieveEmbedding and relevant context
  3. ContinueShared memory across encounters
Models & cost

Explore cloud or local models, APIs, routing and token budgets, with cost limits chosen for the experience.

  1. ModelsCloud or local
  2. RoutingSelect for the task
  3. BudgetControl ongoing use
Sensing & context

AI perception meets MR spatial context through images, video, voice, gestures and wearable signals.

  1. SenseVoice, vision and body context
  2. InterpretAI perception
  3. RespondConversation and MR behavior
Tools & action

MCP tools, proactive messages, app workflows and connected devices link intent to useful actions.

  1. ConnectCalendar, messages, apps and home
  2. CoordinateTool use and intent
  3. ActReminders, check-ins and device control
Build & customize

Bring frontend, backend, deployment, version control and open-source adaptation into a maintainable project.

  1. BuildFrontend and backend
  2. TestInteraction and compatibility
  3. DeployVersioned, adaptable projects
Original framework

A Monstar initiative

A place for
independent creators.

The Creator Program supports people building AI companions, AI + MR experiences and interactive characters, with help presenting, publishing and maintaining their projects.

Bring your idea.
Let’s find its place.

  • Character apps and AI companions
  • Character cards, plugins and interactive ideas
  • Projects with a working version or watchable demo
What support is available?

Project presentation, version information, publishing updates and community support. The scope of each collaboration is agreed together.

How do I get involved?

Contact the team with a brief introduction, project screenshots and a working version or demo link. Ownership, authorization and cooperation terms are discussed before publishing.

Monstar Interactive

A little world.
A new connection.

Let’s make room for what comes next.

LET’S MAKE A CONNECTION

Let’s get in touch.

Connect with the Monstar team on WeChat to discuss AI digital life agents, interactive worlds or collaboration.

Official Monstar company WeChat QR code

Scan in WeChat, or save the QR code to your phone.

Save QR code+86 176 2171 2986

PRIVACY

Keep it simple.

This website does not collect project submissions or contact information directly. Project collaborations are discussed through our company WeChat. Please share only the information needed for those discussions.

System diagram

Open original

Zoom in to read the details. Scroll or swipe to explore.