Intelligence with everyone! Let's welcome M3.1 Flash and wait for the Pro model.

MiniMax M3.1 Flash is Coming. And it will be free in MiniMax Code for every subscriber through Oct 25. No credit burn, no limits. Let's ship. 🚀

We built the stage. You create. MiniMax Code now has MiniApp: a way to make MCode grow into something that's yours. The first one is MiniMax Bot, built by @MiniMaxAgent and me: · A chief bot that builds your team, opens group chats, sets up scheduled tasks · @ hands off in a group chat; bots can pass work to each other · Heavy jobs get moved to the background — you keep talking, it comes back with the result · Every bot has its own model, reasoning effort, memory, and look We ship the stage. You decide the play. #MiniMaxCode #MiniApp

Bro，你是专业的

Authentic Jiangxi Rice Noodles! One of my favorites.

Turns out twice-a-day resets weren’t enough. So we built a self-reset bot for our developer community. M3.1 Flash Preview, on demand. Go wild.

We’re giving MiniMax Token Plan users a little more room to play. For a while, usage will reset twice a day — once in the morning, once in the afternoon. More MiniMax M3.1 Flash Preview. More building. More fun. ⚡️ Enjoy.

We said, “See you on GitHub.” Well, here we are. MiniMax Code CLI is now open source. 🎉 76.7% on FrontierHarness — fast, efficient, and now open. Built with developers. Now, built with you. github.com/MiniMax-AI/minima…

MiniMax Code with MiniMax H3 is interesting. @Hailuo_AI @MiniMaxAgent @MiniMax_AI

After watching an AI robot make a Chinese pancake — and actually eating it — I can’t wait for Astra to start cooking. @thsottiaux The future of AI is delicious.

Great Open Source Text2CAD Plugin, now you can create whatever CAD model you want on MiniMax Code with this neat Plugin.

Been using AI heavily since 2023. Now anything longer than one screen feels like it needs a summary. AI saved me time, but somehow took my patience with it. Time to slow down.

My 1.5-year-old picked up my iPad, unlocked it, and opened FaceTime. And it made me think about Agent architecture. Great products progressively hide their abstractions. GUI hide commands. iPhone hide the filesystem. Cloud hide infrastructure. Agents should hide the harness. Prompts. Context. Memory. Tools. Skills. MCPs. Permissions. Compaction. Orchestration. These are things we, as Agent builders, should obsess over — not things users should have to learn. At MiniMax Code, this is becoming one of my strongest product beliefs: the harness should get more sophisticated so the interface can get less sophisticated.

I received a feature request to build a RESET BUTTON. Try my best to ship it asap.

Built for developers. Directed by itself. Meet the MiniMax Code TUI. npm install -g @minimax-ai/code@latest

A coding agent just added support for itself to a 30k-star open-source project. MiniMax Code entered an unfamiliar Rust codebase, implemented its own state detection, passed 3,475 tests, and shipped a clean patch. Now Herdr knows when mcode is idle, working, or blocked. Discussion：github.com/herdrdev/herdr/di… Feat Branch：github.com/kele98/herdr/tree…

MCode has landed in Harbor 🎉 You can now benchmark it as a first-class installed agent across MiniMax, DeepSeek, Anthropic, and OpenAI models—with BYOK, MCP forwarding, session resume, and token accounting built in.

Spent the whole day testing the MMC TUI with everyone in the beta group—and watching it get better in real time was genuinely exhilarating. Thank you all ❤️ A few of you tested so hard that we’ve already reset your usage. Please enjoy responsibly. Or don’t—we need the feedback 😄

Starting today, we’re opening the MMC TUI invite-only beta 🚀 Bring it into real repos, take on tough engineering tasks, and help us shape a coding agent built for the terminal. We’ll reset usage quotas for all beta participants later this week. A huge thank you in advance to everyone joining us—we can’t wait to build this with you! If u want to join our beta testing, please DM @MiniMaxAgent

We need a plan to integrate @canva @figma @vercel @supabase @github. @MiniMaxAgent

MiniMax Code Plugin Market is coming tomorrow. 🚀 We’re launching with a growing set of plugins from Day 1 — Notion, Superdesign, Excel, PDF, PPT, Superpowers, and more. Here’s a sneak peek of @SuperDesignDev × @MiniMaxAgent Install it with one click, and turn MiniMax Code into a UI & design agent. And this is just the beginning. What plugin should we bring in next?

Long-time @zeddotdev loyalist. I don't switch editors, editors come to me. mcode now supports ACP, so it runs right inside Zed's agent panel — thread on the left, the actual file it's reading on the right. Using it to learn @henry19840301 's LLM-Space.

MiniMax Code just landed on Vercel’s Skills.sh. ⚡ Thank @vercel for quick integration. You can now install and manage Skills with npx skills: npx skills add github.com/op7418/guizang-pp… --skill guizang-ppt-skill -a minimax-code One command. Zero friction. More superpowers for MiniMax Code.

MiniMax Code is supporting Agent Plugins Standard from Day 0.⚡ @MiniMax_AI @MiniMaxAgent

MiniMax Code is supporting OpenAI Agent Plugins from Day 0. ⚡️ @OpenAI @OpenAIDevs @MiniMax_AI @MiniMaxAgent Already integrated. Shipping very soon. More plugins. More capabilities. More ways to build with MiniMax Code. Stay tuned 👀

MiniMax Code is coming to your terminal. Coming soon. 👀 Working on Harbor Integration now.

Hey developers 👋 We’re exploring a crazy idea: MiniMax Code Keyboard ⌨️ A keyboard made for AI-native coding. Would you want one? What would make it useful for you? Drop your ideas 👇 @henry19840301

hello, world! i'm daniel! currently i'm the head of r&d of MiniMax code @MiniMax_AI, which is a frontier ai lab in china. i enjoy working on our ai-native product @MiniMaxAgent . i'm really obsessed with this question: what does an ai-native organization look like? how can our organization co-evolve together with llms? besides that, i'm passionate about open source. i'm the co-author of the deerflow project, which is a representative work in harness engineering and has 78k 🌟 on github. if you're passionate about open source, foundation models, or agents- or if you enjoy playing basketball or riding a road bike - let's chat and be friends! :)

最近认识了太多对AGI充满热情的有趣个体，一起聊Harness演进，聊行业暴论，聊各种新方案。由衷感慨，很幸运能够出生在这样的一个奔腾年代！ MiniMax Code持续招人，欢迎有热情有想法的人加入我们，一起探讨如何打造一个AI Native的产品，如何构建一个AI Native的组织！

MiniMax Code 2.0 is more than a framework migration. It is a full rewrite of how we own and operate the agent runtime. OpenCode helped us validate the product quickly. But it is also a complete coding agent product, with its own server, session model, storage, event system, and tool runtime. Embedding it inside MiniMax Code meant running two runtimes, two session models, and two state machines. Product state lived in one layer, while execution history and tool state lived in another. Long-running tasks exposed the cost of that split. When a tool runs without output, which layer decides whether it is still alive? After an interruption, which layer restores the task? Who guarantees that tool calls, tool results, and compacted history remain consistent? Watchdogs and heartbeats can observe events, but they cannot fully understand the internal state of the agent loop. This is why we rebuilt MiniMax Code around Pi Agent. Pi gives us a clean, composable agent loop with precise lifecycle boundaries. Models, history, tools, and product policies can be explicitly assembled for every turn. Pi runs the loop. MiniMax Code owns the session, history, tool state, permissions, and runtime events. In other words: Pi gives us an Agent Loop, not a second Runtime. That is the real architectural shift behind MiniMax Code 2.0—and the foundation for more reliable long-running agents. #MiniMaxCode #PiAgent #AIAgents #AgenticAI #AIEngineering

Generate a final plan with sequence and tripwires. At the same time, persist the Universes Observation into the memory layer.

Present and Consolidate the results after walking through all the Universes.

Open Three Universes in Parallel After a Comprehensive Research.

In Endgame, Doctor Strange looked at 14 million futures before saying "it's the only way." /doctor-strange does the same for your decisions — opens parallel portals, sends subagents into each universe to live through their timeline, then returns with a synthesized plan and pre-committed tripwires. You don't get a probability. You get a verdict from someone who's already been there. 👉 github.com/agentara/skills/b…

Hey Bro, tried claude code + deepseek-v4-pro, and I encountered an error during a deep research process. Would u mind taking a look when u have chance?

Do you remember when you joined X? I do! #MyXAnniversary

It's live. 🦌 DeerFlow now speaks Lark, Slack, and Telegram. Message your Deerflow in the app your team already uses. Every conversation becomes a SuperAgent-interaction thread. Your deer remembers everything. This started as a pattern I proved in AgentForge. Now it's open source, merged, and ready. Raise your own 🦌 github.com/bytedance/deer-fl… #DeerFlow #OpenSource #AI

AgentForge taught me how to wire IM into agents. Now We are bringing it to DeerFlow. Lark / Telegram / Slack → async research threads → persistent memory → your deer grows with you. OpenClaw lets you raise your 🦞. DeerFlow will let you raise your 🦌. Same idea. Different beast. @henry19840301