Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-21 · high
Alibaba says 15 internal agents already handle 85 percent of its developer support tickets, unveiled at WAIC 2026 in Shanghai.
Why now: Alibaba Cloud made the announcement at WAIC 2026 in Shanghai, per the company's blog post dated for the event.
Alibaba Cloud unveiled an 'Agent Native Cloud' architecture at WAIC 2026 in Shanghai, per the company's blog. Alibaba says 15 coordinated agents already field 85 percent of its own developer support requests. The same agents cut operational support time 90 percent and compressed release cycles to a single day.
The stack is built on native sandboxes, workload isolation, elastic scaling, and enterprise identity integration, layered onto Alibaba's existing AgentRun lifecycle platform. The isolation bet echoes other recent coverage of enforcing agent policy at the kernel level with eBPF. It adds AgentLoop for real-time tracing and evaluation, AgentTeams for multi-agent coordination and governance, and Agentic Computer for secure cloud execution.
Alibaba is framing this as a different bet than AWS AgentCore or Microsoft Foundry's hosted-agent platforms: real production numbers instead of a roadmap promise.
The blog post doesn't say which agents replaced which support workflows, or what the 85 percent figure excludes. If you're picking a cloud for agent workloads, the real test is whether the vendor runs its own agents in production at this scale.
Each link below shares sources, entities, or timing with this story.
At WAIC 2026, Alibaba unveiled a rearchitected stack on native sandboxes, workload isolation, elastic scaling, and enterprise identity, adding AgentLoop (tracing, evaluation, optimization) and AgentTeams (multi-agent coordination) to AgentRun. The proof point is internal and s...
Pricing runs on four model tiers instead of per seat, and the platform slots into DingTalk's 20 million business accounts.
MCP's new roadmap and A2A's move under the Agentic AI Foundation both point at the same target next: agent identity for enterprise deployments.
Repos and secrets stay on your network, but prompts and results still leave for inference, and one user locks each runner.
CloudWatch Coding Agent Insights tracks tokens, cost, and cache hit rate for Claude Code, Codex, and Copilot without extra setup, per AWS.
Three days running, Product Hunt's leaderboard has favored tools that run agents over agents that do anything themselves.
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.