DingTalk Open-Sources DWS, Brings 1,000+ Commands for AI Agents to HarmonyOS PC
DingTalk has released DingTalk Workspace CLI (DWS) for HarmonyOS PC and opened its source code. Enterprise messenger features have been turned into standardized command-line commands that can be invoked not just by humans, but by AI agents too. The scale is notable: 25 business modules and more than 1,000 commands. An agent gets access to group chats, meeting transcripts, tasks, AI tables, documents, calendar, address book, and approvals, with no need to write a separate integration for each service. • Target platform: HarmonyOS PC, Huawei's desktop system • The repository is open, so third-party agents can connect to DWS • Use cases include calendar, AI tables, and approval workflows
AI-processed from Jiqizhixin (机器之心); edited by Hamidun News
The DingTalk Open Platform has released DingTalk Workspace CLI, abbreviated DWS, in a version for HarmonyOS PC, and simultaneously open-sourced the project. The full product functionality of DingTalk is packaged into a standardized command-line interface: 25 business modules and more than 1,000 commands. Through it, office operations can be carried out by AI agents, not just by humans through a graphical interface.
A messenger as a set of commands
Classic enterprise software was built for a human with a mouse: buttons, forms, pop-up windows. All of that is awkward for an agent. What it needs are clear calls with predictable input and output. A CLI provides exactly this format: command, parameters, result. Agents built on large language models are good at using the command line, because there's a huge amount of it in training data.
DingTalk has followed this path and broken its products down into commands. The same layer can be called by a developer's script, an automation, or an autonomous agent. The interface is shared, while whoever connects to it defines the permissions and use cases.
What exactly is available to an agent
The company names the main office scenarios that DWS covers:
- group chats — reading and sending messages in work groups;
- meeting transcripts (会议听记) — access to recordings and text of past calls;
- tasks (待办) — creating and tracking to-do lists;
- AI tables — working with structured data in DingTalk tables;
- documents — creating and editing work materials;
- calendar — scheduling meetings and events;
- address book — searching for colleagues and contacts;
- approvals (审批) — launching and processing corporate requests.
This is only part of the 25 modules, but it illustrates the logic. An agent can take a meeting transcript, extract action items, create tasks, enter data into a table, schedule the next call on the calendar, and send a summary to the chat. Previously, such a chain required several integrations; now it's a sequence of commands in a single environment.
Why the link to HarmonyOS PC matters
HarmonyOS PC is Huawei's desktop system, which the ecosystem is actively filling with applications. For a platform built from scratch — especially one designed for a local market — it's not just an app base that matters. Tools through which agents will operate on it are also needed. DWS fills exactly this niche: DingTalk's office capabilities are available on the system in a machine-readable form.
For DingTalk, this is a way to claim a place in the new agentic infrastructure on yet another platform. If an agent already knows how to work with corporate processes through DWS, it becomes easier for it to gain a foothold within a company. For owners of HarmonyOS PC devices, this is a ready-made set of actions that can be plugged into an assistant without developing connectors from scratch.
Open source code and a bet on third-party agents
Open-sourcing the code changes the product's role. DWS becomes a shared interface that developers of agentic systems can embed into their own solutions, rather than a closed DingTalk feature. The code can be studied, adapted to specific tasks, and checked to see exactly which permissions an agent receives when calling a given command. For enterprise customers, this last point is often decisive: giving a model access to chats, documents, and approvals is unsettling without transparent boundaries.
That's why access separation and action logging are critical in such projects. Approvals and contacts contain sensitive data, and a mistake by an agent here costs more than an inaccurate answer in a chat.
Where developers and IT departments should start
A sensible starting point is a pilot on low-risk tasks: meeting transcripts, tasks, and calendar. Results are easy to verify, and the cost of an error is small. Only after that is it worth connecting documents, tables, and approvals — with separate permissions for the agent, rather than under a live employee's account.
Those building their own agents would do well to look at the number of commands as a ready-made map of capabilities. More than 1,000 calls mean there's no need to invent an integration layer, letting developers focus on logic: what the agent should do, when, and with what confirmation from a human. Companies on HarmonyOS PC should assess which routine office chains can be assembled from DWS commands right now.
- Oct 27, 2026Fudan and Huawei Xiaoyi Teach an AI Agent to Write Operators for Ascend NPUs: 95.5% Success Rate
- Oct 27, 2026EmbodiedLab: robots tested on experiments, and the best one solves less than half
- Oct 27, 2026Skill∞: an agent improves its own skills and its own meta-skill, 60.7% on SkillsBench
- Oct 27, 2026Work Agents: 16 Researchers Build a Unified Coordinate System for AI Agents
Key Questions Answered
What are the key details on A messenger as a set of commands?
Classic enterprise software was built for a human with a mouse: buttons, forms, pop-up windows. All of that is awkward for an agent. What it needs are clear calls with predictable input and output. A CLI provides exactly this format: command, parameters, result. Agents built on large language models are good at using the command line, because there's a huge amount of it in training data. DingTalk has followed this path and broken its products down into commands. The same layer can be called by a developer's script, an automation, or an autonomous agent. The interface is sha
What exactly is available to an agent?
The company names the main office scenarios that DWS covers: group chats — reading and sending messages in work groups; meeting transcripts (会议听记) — access to recordings and text of past calls; tasks (待办) — creating and tracking to-do lists; AI tables — working with structured data in DingTalk tables; documents — creating and editing work materials; calendar — scheduling meetings and events; address book — searching for colleagues and contacts; approvals (审批) — launching and processing corporate requests. This is only part of the 25 modules, but it illustrates the logic. An
Why the link to HarmonyOS PC matters?
HarmonyOS PC is Huawei's desktop system, which the ecosystem is actively filling with applications. For a platform built from scratch — especially one designed for a local market — it's not just an app base that matters. Tools through which agents will operate on it are also needed. DWS fills exactly this niche: DingTalk's office capabilities are available on the system in a machine-readable form. For DingTalk, this is a way to claim a place in the new agentic infrastructure on yet another platform. If an agent already knows how to work with corporate processes through DWS,
Delegate Deep Research, Coding & Workflows to an AI Agent
Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.