UnitySight · 一元洞见

Multi-agent. Multi-app. Multi-device.
The foundation for vertical agent apps.

UnitySight builds infrastructure for vertical agent applications — orchestrating multiple agents, connecting multiple apps and spanning multiple devices, so agents run reliably and under control inside real workflows.

01Platform

One foundation for every form of vertical agent app.

A layered design with clear responsibilities and loose coupling. Security and governance span every layer.

UNITYSIGHT · Reference architectureFIG. 01
L4Applications
Productivity assistant
Developer apps
Agent app store
L3Orchestration
Multi-agent coordination
Task planning
Capability routing
L2Capabilities
Tool & app integration
Memory & context
Model-agnostic
L1Data & Sync
Structured personal data
Cross-device sync
Local-first
L0–L4Governance
Permissions
Privacy
Auditability
L0Surfaces
Desktop
Mobile
Web
Terminal
MULTI-AGENT

Multi-agent

Composable · Collaborative

A unified orchestration layer coordinates multiple agents and routes each task to the right model and capability, with no single-vendor dependency.

MULTI-APP

Multi-app

Unified integration · Open ecosystem

Apps share consistent data and context. Third-party developers integrate through the same standard, and capabilities are reused across apps.

MULTI-DEVICE

Multi-device

Continuous · Local-first

Tasks and state move seamlessly across desktop, mobile, web and terminal, with data kept on the user's side by default.

Model-agnosticPrivacy-firstExtensibleAuditable

02Why now

Models have matured. The bottleneck is deployment.

An agent's value no longer depends on the model alone, but on whether it can enter a specific context, retain that context over time and complete tasks reliably.

Capability inflection

Reasoning, planning and tool use are now sufficient for a wide range of real tasks.

Form-factor mismatch

Chat interfaces cannot carry use cases that require durable state and structured data.

Vertical gap

Every domain has distinct processes, rules and data, and needs purpose-built agent applications.

03The problem

Building vertical agent apps still faces four systemic challenges.

01

Fragmented context

Data is scattered across apps and devices, so agents never see the full picture.

02

Costly workflow fit

Generic agents lack domain understanding; integrating real workflows requires heavy customization.

03

Insufficient privacy and control

Personal and business data must be processed in controlled, auditable environments.

04

No shared foundation

Developers repeatedly rebuild orchestration, memory, integration and sync.

04Market

We enter through niche vertical user groups that general platforms overlook.

General platforms are built for the mass market and cannot design deeply for every segment. These groups have well-defined needs and high-frequency workflows — where vertical agents deliver the most value.

Well-defined needs

Clear, recurring workflows that agents can take on end to end.

Structural gap

General products cannot cover every niche, leaving a structural gap in supply.

Depth, then breadth

Go deep in one use case, then expand to more verticals through the open platform.

05Roadmap

One foundation, three stages, from product to ecosystem.

Each stage reuses and strengthens the same foundation.

Stage 1In development

Personal productivity assistant

Manage todos, notes and expenses in natural language, with linked data and context that compounds over time.

TodosNotesExpenses
Stage 2Planned

Developer platform

Open the foundation to developers to connect their own apps and build and ship vertical agent applications efficiently.

App integrationReusable capabilitiesVertical use cases
Stage 3Future

Agent app store

Developers publish, distribute and monetize their apps; users discover the agent applications that fit their needs.

PublishDistributeMonetize

06UBot

UBot: a local-first AI assistant on your own devices.

UBot is a local-first AI assistant. Each bot has its own workspace to work with files, run commands you approve, browse the web, and follow routines, on Mac and phone, with your data staying on your devices. Built for personal and household use, and as a base for focused apps.

Local-firstPer-bot workspaceCommands you approveRoutinesMac · PhoneGroup collaboration
UI illustration · Sample dataDrawn from the actual apps with sample data. The UBot Apps tab is coming soon.
9:41
Chats
Search
🏠
FamilyGroup09:12
Life assistant: Saturday shopping list updated2
🌿
Life assistant08:30
Weather reminder set for 8 AM tomorrow
💰
Budget assistantYesterday
This week's spending is in a table
💻
Code assistantMacYesterday
cargo build finished with no errors
CC
Claude CodeExternalMon
Via makit · Waiting for your approval1
📚
Study assistantMon
Reading notes saved to Documents
Chats
Apps
Devices
Me
9:41
Apps
Family apps
微行动Build habits bit by bit; redeem points for rewards
合拍Move tasks forward with AI: hand off, then review
mindflowCapture ideas anytime; organized into notes automatically
+
More appsDeveloper integration
Opening soon
Apps run inside UBot and share its assistants, devices and permissions.
Chats
Apps
Devices
Me
9:41
Devices
Connected
💻
My MacBookmacOS · Direct
Online
📱
iPhoneThis device
Recent activity
💻
Code assistant ran a command on Maccargo build
2 min ago
🌿
Life assistant read a fileshopping-list.md
10 min ago
💰
Budget assistant routine ranWeekly bills summary
1 hr ago
CC
Claude Code command approvedgit status
Yesterday
Chats
Apps
Devices
Me
9:41
Me
🙂
Sample user1 Mac · 1 iPhone
💻
Devices
2
🧠
Models
Auto
🔗
Connections
makit · Cursor
🔒
Privacy & Security
Ask me
⚙️
General
Chats
Apps
Devices
Me

Use cases

Existing capabilities, plus integrations with apps coming soon.

Family

Today, each family member joins UBot as a device. The multi-device foundation lets several people and several assistants share one conversation and one plan.

🏠Family
👩
Mom
Let's split the chores this week
🌿
Life assistant
Here's this week's rota, based on everyone's schedule:
Mon
Take out trash
Leo
Wed
Groceries
Dad
Sat
Deep clean
Everyone
⏰ Reminders at 8 PM · Check in with 微行动 when done
✓Leo finished "Take out trash"
F1Multi-deviceHousehold choresFamily members and the life assistant split chores in the Family group: an automatic rota, timely reminders and completion tracking.
🏠Family
👨
Dad
Camping next weekend? Any ideas?
👦
Leo
Somewhere with a lake!
🌿
Life assistant
Combined everyone's input. Suggested plan:
Leave Saturday · Lakeside campsite · 1 night
Book campsite
Dad
Buy food
Mom
Pack gear
Leo
Synced to 合拍
F2Multi-deviceNear-term plansThe family plans a weekend or next month together with the assistant, which gathers everyone's input, proposes a plan and breaks it into todos.
Every day at 8 AM, summarize my schedule and the weather
⏰
Morning briefingDaily 08:00
01RoutinesRecurring tasks, done on schedule.
💻Code assistant wants to run on "My MacBook"
$ cargo test --release
DenyApprove
02Command approvalNothing runs until you approve it.
🏠Family
🌿
Life assistant
Saturday shopping list is ready: 8 items.
💰
Budget assistant
Estimated the cost and added it to this month's budget.
03Group collaborationSeveral assistants split the work in one group.
Connections
makit
Connected
Claude Code
Via makit
Codex
Via makit
Cursor
Connected
04External assistantsConnect Claude Code, Codex and more through makit.
My MacBook●
ActivityFilesTodos
$ cargo buildDone
$ git pullDone
$ npm testAwaiting approval
05Control your Mac from your phoneSee your Mac's activity, files and todos on the go.
🌿Life assistant wants to read a file
Documents/family-bills.csv
Allow onceAlways allowDeny
06Privacy, ask firstUBot asks before touching files; data stays local.

Family apps

Focused apps built around UBot. Each app can sync its data to UBot assistants, so they can help based on your habits, tasks and notes.

微行动

Build habits bit by bit; redeem points for rewards

Habit check-insPoints & rewardsStats & heatmap

Syncs data to UBot assistants

9:41
HabitsBuildQuit
Tue, Nov 10
3/41 more for a perfect day
Points128
AllAMPMEveningTodo
🌅
Wake up early12-day streak
✓
📖
Read 30 min8-day streak
✓
🏃
Workout3/4 this week
✓
🧘
Meditate5-day streak
Habits
Rewards
Stats
Me
9:41
Stats
Completion76%
Check-ins86
Last 8 weeks
Best time7–9 AM
Strongest dayWed
Last 6 months
Habits
Rewards
Stats
Me

合拍

Move tasks forward with AI: hand off, then review

Hand off to AIReview resultsAI work stats

Syncs data to UBot assistants

9:41
For meUBot connected
2 to review
Prepare next week's pitch deckTo review
Done by AI · 12-slide outline & talking points
AcceptReturnView
Compare three cloud quotesTo review
Comparison table · incl. annual discounts
AcceptReturnView
Book a room for Friday PMAI doing
Checking available slots…
New task…Give to AI
For me
AI doing
Lists
Me
9:41
Me
14Done this week
9By AI
6.5hAI saved
This weekYouAI
M
T
W
T
F
S
S
BotUBot
UBot connected
For me
AI doing
Lists
Me

mindflow

Capture ideas anytime; organized into notes automatically

Quick capture#tagsDaily notes

Syncs data to UBot assistants

9:41
Capture
9 this week · 4-day streak
09:20

Product idea: link todos and notes into one timeline

#product
11:05

Open the pitch with a real user story

#pitch
21:10

☑ Send homepage feedback to the designer tomorrow

#todo
What's on your mind…
#☑[[ ]]Supports #tags
Capture
Notes
Settings
9:41
Notes
AllDaily#product#reading
Nov 10 · Daily note4 ideas · 1 todo
Product idea#product Link todos and notes into a timeline…
Pitch prep#pitch Open with a user story, then the plan…
Local-first software#reading Data stays on your devices by default…
Capture
Notes
Settings
WAITLIST

Join the UBot waitlist

Leave your email and we will let you know as soon as early access opens.

We only use your email for early-access and product updates.

07Shipped

We already ship.

makit is our released developer tool, and the environment we use every day to build our agent products.

Released · Open source

makit

A high-performance terminal for AI coding sessions. Run Claude Code, Codex and OpenCode side by side in splits and tabs.

Type
Developer tool
Built with
Native Rust
Runs
Local only
Works with
Claude Code · Codex · OpenCode

08Partnerships

Partner with us to bring vertical agents into production.

We welcome collaboration in the following areas.

CO-BUILD

Vertical co-development

Work with industry partners to define requirements, refine workflows and jointly build vertical agent applications.

DEVELOPERS

Developer & app integration

Connect your own apps to the open platform, join early access, and help shape the foundation and developer experience.

CHANNEL

Channel & ecosystem partnerships

Work with channel, platform and ecosystem partners to reach more niche user groups and vertical use cases.

STRATEGIC

Investment & strategic partnerships

Conversations with investors and strategic partners about our direction, product roadmap and long-term plans.

For business inquiries, contacthi@unitysight.com →

Talk to us.

We welcome conversations with investors, developers and partners. Progress is published openly on GitHub.