Teamily AI, the Silicon Valley startup building a messaging platform where AI agents participate as named members of group conversations, has officially launched in Singapore with native iOS and Android apps. The company is backed by $20 million in seed funding raised through its parent entity, TensorOpera AI, according to a PR Newswire announcement published May 26.
The company was co-founded by Dr. Aiden Chaoyang He and Prof. Salman Avestimehr, who first met at USC where He was Avestimehr’s PhD student from 2018 to 2022. He previously worked at Tencent, Meta, Google, and Amazon. The two originally co-founded TensorOpera AI, a generative AI developer platform, before spinning out Teamily as a consumer-facing subsidiary focused on multi-agent group collaboration, as Forbes reported in March.
How Teamily Works
Unlike OpenClaw or Hermes Agent, which operate as solo-agent harnesses controlled by a single user, Teamily treats agent collaboration as a social layer. Multiple AI agents appear as named participants in group chats. Users can mention them, assign tasks, and run parallel jobs while the conversation continues.
The platform runs on three technical layers, according to Forbes: a universal memory layer that retains context across groups and sessions, a “social brain model” planning engine that distributes tasks across agents, and an agent social network where humans and AI agents coexist in real time. In a live demo described by Forbes, agents within a single group chat simultaneously handled market research, slide preparation, and document drafting while human participants continued discussing the output.
Singapore as the Launchpad
The Singapore market entry aligns with the country’s updated Model AI Governance Framework for Agentic AI, published by IMDA in May 2026. That framework introduced new best practices for multi-agent systems, managing risks from third-party agents, and guarding against automation bias. Singapore has positioned itself as a regulatory sandbox for agentic AI, with OpenAI also opening its first overseas Applied AI Lab there earlier this month backed by $234 million in commitments.
Three Models, Three Bets
The agent framework market now has three distinct architectural approaches competing for different user profiles. OpenClaw operates as an open-source, community-driven solo harness for individual builders. Hermes Agent targets developers with a performance-first, self-improving framework. Teamily bets on a third path: collaborative agent teams embedded in social messaging, aimed at groups, families, and professional teams who want multi-agent coordination without touching infrastructure.
Whether group-based agent collaboration can compete with the traction of solo-agent harnesses (OpenClaw crossed 145,000 GitHub stars this month) remains an open question. Teamily’s $20 million gives it runway to find out.