Following Anthropic's recent launch of Claude Tags, which integrates Claude as a team member in Slack, AI visionary Karpathy highlighted that this represents the third paradigm shift in LLM interaction. It transitions LLMs from web interfaces and desktop apps to persistent organizational entities with full team tools and context.
Riding this wave, Kunlun Tech has unveiled Skywork Tags in its latest TianGong 3.2 upgrade. This feature directly connects the #Skywork AI Agent with popular instant messaging tools, including Lark, DingTalk, Slack, Discord, and Telegram, allowing the agent to "work" wherever the team collaborates.
While most AI Agent products on the market require users to painfully migrate scattered chats, documents, and historical data into a new workspace, Skywork Tags flips the script. Instead of forcing teams to move their context, it brings the Agent into their existing workflow. By simply @-mentioning the bot in an active group chat, it immediately accesses the context without requiring extra log-ins or data exports.
A key differentiator of Skywork Tags is its shared architecture. Unlike isolated personal assistants, there is only one Skywork instance per channel. This brings three major benefits: Transparency (everyone sees the agent's progress), Asynchronous Hand-off (team members can pick up where the agent left off), and Continuous Context Accumulation (the agent grows smarter as it observes team interactions over time).
The product's effectiveness was validated by Kunlun Tech's internal experiments over several months. Initially, customized personal bots performed better. However, within two to three weeks, a shared bot deployed in a 100-person group chat—which learned from diverse inputs across product managers, developers, and operators—exponentially surpassed the personal versions. This suggests that collective human feedback is the fastest pathway to scaling an enterprise AI Agent.
[AgentUpdate Depth Analysis] The launch of Skywork Tags represents a pivotal shift from AI as a personal "Copilot" to AI as a "Collaborative Peer." By piggybacking on mature enterprise communication platforms like Slack and Lark, Kunlun Tech avoids the high friction of onboarding users to a new platform—a strategy that holds significant practical value for enterprise AI adoption. In the long term, this shared-agent paradigm challenges the individual-assistant model. It highlights that the next battleground for AI Agent ecosystems is not just model intelligence, but seamless integration with organizational workflows and collective memory. The transition to hybrid human-agent workforces is accelerating, and "in-context collaboration" is the killer app that will define the enterprise GenAI market.
