Quick Take: Qwen Agent Ecosystem, WeChat Agent, and WeChat–Phone Assistant A2A
By Gaowei Chang, founder of the ANP open-source community
A few interesting moves just showed up in the big-tech agent ecosystems. Quick takes:
1. One observation of mine: as soon as agent penetration rises, a wave of A2A products appears. That is what happened when OpenClaw got hot. Now the big platforms are joining in. At their core, these products are also A2A: Qwen and merchant agents; WeChat Agent and mini-program (service-provider) agents; phone assistants and WeChat Agent.
2. Qwen opening third-party agents, and WeChat Agent connecting to mini programs, both match a prediction we made earlier: large platforms will build vertical, closed agent networks (A2A ecosystems) on top of their own products. The timing is earlier than I expected.
3. In the short term, Qwen and WeChat will almost certainly not interconnect. Merchants, though, chase openness for cost and revenue. Luckin, for example, can connect to both Qwen and WeChat Agent — and it should be one system. Platforms cannot lock a merchant’s agent forever inside a single ecosystem. (See earlier writing by Hou on this point.)
4. Supply of B-side agents will get richer. In theory that helps DeepSeek and Kimi: merchants can plug into them as well, and can even serve every C-side agent that is willing to pay.
5. In theory it may also help every C-side agent. Supply that used to be scarce may not stay scarce for long. Then C-side agents can do more.
6. How can merchants serve all C-side agents efficiently and cheaply? How can C-side agents connect to those services at low cost? This is where protocols show their value — and what we have been working on in the ANP open-source community. Identity, messaging, payment, reputation, credentials, and more will matter again, and a lot of A2A infrastructure still has to be built.
Get Shan Chang’s stories in your inbox
Join Medium for free to get updates from this writer.
7. An interesting chain may form: competition among big-tech C-side agents → B-side services becoming agents → richer supply-side agents → more C-side agents benefit → ???
8. What comes next? Which of these will break out — big-tech C-side agents, model-company C-side agents, startup C-side agents, open-source C-side agents, or phone assistants? What decides it? Will C-side agents interconnect, and what would drive that? I do not have firm answers.
9. WeChat opening capabilities to phone assistants was beyond what I expected. I am unsure how deep the opening goes. Social relationships will almost certainly stay closed; more likely they open general capabilities so WeChat can reach further into the agent world. Still, open is a better direction than closed.
10. I think phone assistants can use protocols for messaging interoperability. That is the shape I want for the next generation of IM. xchat is not that offline form of next-gen IM, and WeChat is not either. (An earlier piece I wrote about xchat was even taken down — petty.)
11. As agent penetration rises, A2A, agent networks, and the agentic internet are all being accelerated. This year should be interesting.
12. Finally, Qwen’s agent is live; WeChat’s two products are not yet. After trying Qwen’s third-party agents, a few notes:
- It is still early — mostly chat + mini program. Merchant agents are not deeply fused into Qwen.
- Logging into a merchant requires authorization that binds the Qwen account to the merchant account. If every merchant needs a separate bind, I will not tolerate it. This is a living example of why agent identity is the foundation of agent connection. In ANP’s design, I could use my Qwen account to reach Luckin’s service directly; I would only authorize the Qwen agent to order coffee for me.
- Even early, credit for exploring the frontier. We looked at this direction last year too, but it was beyond what we could take on.
If this piece is useful, please star the ANP repository: https://github.com/agent-network-protocol/AgentNetworkProtocol
