Built for the AI-native enterprise.

Humans and agents. One team at work.

An AI Native workspace for humans and agents.

See the team at work

置顶

队友

IM

今天

Vino

我们准备下周发布 Arkhe 官网。帮我整理一下要做的事,每项写清交付物和验收标准。

Ally

可以,先把发布前的工作拆成三项。

  1. 核对产品界面。逐屏记录实际入口与能力,交付一份界面对照表。
  2. 整理协作场景。写清目标、对话和结果,交付一份演示脚本。
  3. 检查中英文页面。核对术语、链接和布局,交付一份页面检查清单。

验收时逐项对照实际产品,确认文案、界面和操作能够对应起来。

我会把这些内容整理到 官网发布清单.md,后续有调整就继续更新。

Vino

今天先核对真实界面和演示内容,中英文检查放到下一步。

Ally

已调整,今天先完成这两项。

  1. 核对导航、会话列表、消息和输入框,逐项记录尺寸与样式。
  2. 按照实际界面整理演示对话,确认每处展示都有对应依据。

中英文检查单独保留,下一步继续。范围和验收标准已经更新到同一份清单里。

发消息给 Ally…

GIVE EVERY AGENT A ROLE

Every agent,one clear role.

Start with one Ally and grow the team as the work demands.
Give every teammate a clear role, so everything has an owner.

A few roles you can give a teammate.

Ally with clock, folder, magnifier, engineer, and star-shaped AI teammates standing together
01

Ally

Your coordinator

Vague ideas are welcome. Talk it through with Ally and shape what needs doing.

02

Tempo

Project management

Break down milestones and set routines. Put what matters on the calendar.

03

Keeper

Knowledge keeper

Turn scattered material into documents, so today’s thinking informs the next round.

04

Scout

Researcher

Read, compare, and weigh the evidence. Turn information into your own judgment.

05

Chan

Senior engineer

Read code, edit files, run tests. From an idea to a running version.

06

Nova

Product designer

Refine flows and copy so every step of the product feels natural.

Channels

Familiar channels. New teammates.

Manage your agent team on desktop. Assign tasks and add requirements in group chats. Connect DingTalk by scanning a QR code to bring agents into everyday work conversations.

DingTalk · Scan to connectGroup chats · Assign and refine

Manage on desktop · Collaborate in group chats

DingTalk
Arkhe teamInternal4 members · Team chat

Today 09:41

@Tally A new teammate just joined. Introduce Arkhe and what you can do.

Read by all
Tally Bot

Welcome! I’m Tally, your AI teammate in this chat.

Arkhe is a local desktop app. Talk to Ally and give each teammate a role so we can work together.

  1. Assign tasks and refine requirements in chat
  2. Organize notes, record decisions, and track work
  3. Set recurring tasks to run on schedule

Mention @Tally in the group to get started.

@Tally Help us outline the work for this website launch.

Read by all
Tally Bot

Sure. Here are three things for us to track.

  1. Review the product copy and demo conversations
  2. Check desktop, mobile, and both languages
  3. Prepare a launch checklist and record open questions

Add any new requirements right here in the chat.

@Tally Focus on the first two today. Leave the checklist for tomorrow.

Read by all
Tally Bot

Got it. Today we’ll review the copy, demo conversations, and pages across devices.

The launch checklist can wait until tomorrow. I’ll keep sharing progress in this group.

DingTalk group · Simulated conversation
November 2026
GMT+8
Mon9
Tue10
Wed11
Thu12
Fri13
Sat14
Sun15
09:0010:0011:0012:0013:0014:0015:0016:00
14:20
Arkhe calendar · Interactive demo

Schedules

Everyday work,right on schedule.

Morning briefings, daily archiving, weekly reviews. See every bot’s recurring work in one calendar, with a clear view of what’s running and what comes next.

Click any event to explore its schedule and run record.

Projects

Every project carriesits own context.

A project gathers chats, instructions, memory, and reference files in one place. Teammates work around the same project, so discussions stay grounded and results stay traceable.

Select a project on the left to see how it organizes work.

Projects/Website launch

Website launch

Copy, pages, and the release plan for the Arkhe site.

Every conversation in a project shares the same instructions, memory, and files.

Recent activity

Chan fixed the mobile horizontal scroll
Keeper archived 3 discussion notes
Tempo updated the Thursday review plan

Project knowledge builds up with every chat and is shared by all teammates.

Arkhe project space · Interactive demo

Capabilities

Not just chat.

From delegation to execution, the mechanisms human-agent collaboration needs, in one work environment.

Assign by role

The Lead delegates by role, so the right teammate takes each task.

ResearchPlanningCode

Separate workspaces

Information and context stay focused on the work each teammate owns.

FilesContextRun records

One calendar

Every bot schedule in one shared calendar, color-coded by teammate; recurring work runs on plan.

MTWTFSS
Prepare the weekly planEvery Monday · 09:00

Guide mid-run

Add guidance in the conversation; scope updates immediately and runs are fully recorded.

Narrow the scope to these three options.

Scope updated for the next steps

Project context

Instructions, memory, and reference files live with the project.

InstructionsMemoryFiles

Local-first

Identities, context, and run logs stay in local SQLite, with a clear data boundary.

Local storageLocal runtime

Local-first

Work records liveon your machine.

Agent identities, context, and run logs are stored locally. The desktop app works with a local runtime service, so you can manage and back up your work data.

Local storage

Identities, context, and run logs live in a local SQLite database.

Local runtime

The desktop app works with a local service; scheduled tasks run on your machine.

Clear boundary

Model calls and external services receive relevant content according to your configuration.

Your work device
Agent identitiessqlite
Work contextsqlite
Run logslog
Arkhe · Local service

Platforms

Where you work.

Start on the desktop: manage teams and workspaces on your computer. More platforms are on the roadmap.

Desktop

macOS · Windows

Mobile

Check progress, handle approvals and notifications

Web

Open the workspace in a browser

The platform roadmap is evolving — tell us which platform you need first.

Questions & answers

A few things before you start.

What is Arkhe?

Arkhe is an AgentOS for the AI-native enterprise: a work environment for humans and agents. Set goals and define roles. The Lead assigns tasks to agent teammates, and recurring work can run on a schedule.

Can I start with one agent?

Yes. Define one teammate’s role and give it a specific task. When you need more specialization, the Lead can create and manage additional teammates.

How do I adjust direction during a run?

Add requirements or correct direction while a task is running. Execution continues with the additional guidance. Interrupted and unfinished runs retain their actual status.

Where do scheduled tasks run?

They execute through Arkhe’s local runtime service. Keep that service available and provide the sources, model configuration, and access permissions the task needs.

Where is work data stored?

Agent identities, context, and run logs are stored in local SQLite. Model calls and external services receive relevant content according to your configuration.

Built for the AI-native enterprise.

Bring your first agentonto the team.

Define a teammate’s role. Start collaborating on one concrete task.

Self-hostedRuns
On your machineWork data
On scheduleRecurring work
Scale as neededTeams