MiniMax today announced the launch of MaxClaw, a new agent mode on the MiniMax Agent platform designed to make advanced AI agents instantly usable—without deployment friction, infrastructure setup, or tool wiring. MaxClaw is fully powered by MiniMax 2.5, MiniMax's open-source flagship model, and introduces native long-term memory, enabling agents that are powerful, affordable, and persistent.
MaxClaw is available now at maxclaw.ai. Built end-to-end on MiniMax 2.5, an open-source model recognized for strong reasoning and low inference cost, MaxClaw offers key advantages: open-source foundation for transparency, lower operating cost compared to proprietary alternatives, and instant cloud deployment without local setup.
A defining feature of MaxClaw is its built-in long-term memory system. Unlike stateless assistants, MaxClaw agents can remember user preferences and working styles, retain context across interactions, accumulate knowledge from past tasks, and improve output consistency without repeated prompting. This transforms MaxClaw into a persistent digital collaborator that grows more effective over time.
MaxClaw delivers a genuine one-click experience for the OpenClaw ecosystem. From the MiniMax Agent web interface, users can activate a fully configured agent without manual deployment. For users who don't need OpenClaw, the platform provides over 10,000 ready-made expert agents covering software development, creative writing, financial analysis, and market intelligence. Users can also create custom agents, publish them to the Expert community, and earn platform rewards.
MaxClaw comes preloaded with production-ready skills, including image generation, video generation, web search, and application deployment. It integrates directly with collaboration platforms such as Feishu and DingTalk, allowing tasks to be assigned through everyday chat interfaces. From clicking "Create MaxClaw" to a running agent gateway takes under 20 seconds.
In real-world use, MaxClaw interprets natural language instructions, resolves relative time references, executes multi-step reasoning, and returns analysis and generated assets in a single flow. With long-term memory, built-in skills, and MiniMax 2.5 at its core, MaxClaw represents a shift from experimental agents to reliable, persistent AI workers.
MiniMax builds open, scalable AI models and agent systems focused on real-world execution, aiming to move AI beyond conversation toward memory-enabled, action-oriented intelligence.
