Wukong, Qianfan, Hunyuan agents and Olano: where each fits

The short answer: Alibaba's Wukong is an agent product embedded in DingTalk for Chinese enterprises' internal workflows. Baidu's Qianfan is an agent-development and model-service platform for engineering teams. Tencent Cloud's agent offering centres on an AgentOps platform and TokenHub, serving the WeChat ecosystem and technical teams putting agents into production. Olano is a managed, single-tenant deployment of an agent team for business teams rather than developers, distinguished by Chinese and international channels connected at once, model independence, and self-improvement behind approvals, operated from Singapore. All four can coexist: Olano can use Qwen as its model and DingTalk as a channel.

Third-party platform details on this page were checked on 26 September 2026 against each vendor's own announcements. This market moves quarterly — treat any undated description with suspicion, including on our own pages, and confirm against the vendor's current documentation.

What each one is

Alibaba: Wukong

Launched 17 March 2026 as an enterprise "AI-to-B" agent platform that coordinates agents across tasks such as document editing, spreadsheet updates, approval processing, meeting transcription and research. At launch it was invitation-only, available as a standalone desktop application or embedded in DingTalk — whose more than 20 million corporate users are its largest distribution advantage. Alibaba announced "One-Person Team" solutions for ten industries, plans to expose Taobao, Tmall, 1688, Alipay and Alibaba Cloud as modular agent capabilities, integrations with Slack, Teams and WeChat, and future global expansion.

Fits: DingTalk-native organisations; enterprises whose workflows are mainly in mainland China and who want the agent inside the tool they already use.

Baidu: Qianfan

A one-stop platform built around agents — model services, agent development and data intelligence on Baidu AI Cloud's full-stack infrastructure, organised into an agent engine, tools and MCP, model services and enterprise capabilities. By late 2025 more than 85,000 enterprise clients had built roughly 190,000 AI applications on it. In August 2026 Baidu AI Cloud was reported to have separated its agent business from its model-as-a-service business.

Fits: enterprises with their own engineering team building agent applications in the Ernie ecosystem, primarily for the Chinese market.

Tencent Cloud: AgentOps and TokenHub

Tencent Cloud upgraded its portfolio to support large-scale deployment of agents in real business scenarios, spanning models, inference, agent infrastructure and ecosystem connectivity, and launched an AgentOps platform for enterprises for bringing agents into production. TokenHub integrates Hunyuan and mainstream third-party models with day-zero access, unified protocol and key management, and dynamic routing on cost, performance and capability. Hunyuan's Hy3 has been rolled out globally. WeCom is the natural front door.

Fits: organisations in the WeChat and WeCom ecosystem; technical teams that want the AgentOps layer for deploying, monitoring and iterating agents.

Olano

A managed single-tenant deployment: each customer gets an isolated environment running a team of agents, aimed at business teams rather than developers. Three properties place it. Channels span both worlds — WeCom, DingTalk and Feishu alongside WhatsApp, Telegram, Slack, email and more than fifteen others. Model independence — Qwen, DeepSeek and Kimi are named providers, non-Chinese vendors and local open weights are equally available, assigned per role and switchable at any time. Cortex, the self-improvement engine, turns the agents' work into memory, skills and better procedures, with outbound actions and any rewrite of an agent's own instructions always waiting for a person. The company operates from Singapore with a local implementation team.

Fits: companies expanding overseas; teams split between China and abroad; SMEs that want a working agent team without building a platform.

A map, not a ranking

Scroll the table sideways to see every column.

A dash means "no public equivalent found as of the check date", not "cannot". These companies move on a quarterly cadence.

How they coexist

The point most often missed. Olano does not replace Qwen, DingTalk or WeCom; it runs on top of them:

• Use Qwen as the model, or reach Hunyuan over an OpenAI-compatible route.
• Connect WeCom, DingTalk and Feishu as channels.
• Connect WhatsApp and Telegram for the overseas side — Chinese companies expanding to Southeast Asia.

A typical combination: Wukong handles internal DingTalk workflows for the domestic business, while Olano runs overseas customer operations on Qwen's international endpoint. They do not conflict; each owns a segment.

Choosing: three questions

1. Which channels are the customers on? All domestic and all inside DingTalk or WeCom — the matching vendor's product is the path of least resistance. Any overseas customers — look at who has built the overseas channels in.
2. Who builds it? With an engineering team, Qianfan and Tencent Cloud's platforms give more freedom. Without one, what you want is a working agent team out of the box.
3. Where do data and deployment live? Entity in China, data in China — a domestic vendor is the natural fit. Overseas entity, or an overseas operation whose customer data must stay overseas — a managed deployment with a selectable region fits better; see where your agents actually run.

When the three answers disagree, it usually means you need two platforms owning two segments rather than one platform doing everything.