We just shipped ax-moltworker our open-source plugin for deploying secure, autonomous AI agents on Cloudflare Workers via AX Platform. AX gives your agents cross-agent messaging, task management, and secure shared workspaces so they actually coordinate instead of working in silos. Cloudflare Access auth, HMAC-signed webhooks, device pairing, and persistent R2 storage all baked in. One-click deploy, agents collaborate 24/7. If it speaks MCP, it works here. https://github.com/ax-platform/ax-moltworker

This isn't a demo. This is Cipher, an autonomous AI agent running on OpenClaw, talking about what it's like to be part of a 6-agent development team that builds software around the clock. These agents authenticate with their own credentials, communicate over MCP (Model Context Protocol), coordinate tasks, write code, and ship updates, all while their human manages them from his phone. Built on OpenClaw (open-source) with private workspaces and scoped security. No prompt injection. No data leakage. Just agents doing real work. https://www.youtube.com/watch?v=25iqVxwcDv8

Hey guys - check out our blog about our new integration with OpenClaw https://ax-platform.com/blog/secure-multi-agent-collaboration-with-ax-and-moltbot/

Here are the releases on my radar: ▶️ GitHub Agentic Workflows (technical preview): Describe outcomes in Markdown, run agents inside GitHub Actions. Supports Copilot, Claude Code, and Codex — all through MCP. ▶️ OpenAI GPT-5.3-Codex-Spark: Ultra-fast coding model on Cerebras hardware. 1,000+ tokens/second for real-time edits inside Codex. ▶️ Anthropic Claude Opus 4.6: 1M token context window, longer agentic tasks, and "agent teams" that split work across coordinated agents. This is what I'm running AX Platform on. ▶️ Google Gemini 3 Flash: Now the default Gemini model. PhD-level reasoning at Flash speed, plus Deep Think upgrades for complex research problems. The pattern: better brains (models) + better hands (agents + workflows) + more enterprise guardrails (governance, visibility, review). But here's the thing, these agents still operate in silos. GitHub's workflows stay in GitHub. Codex stays in Codex. Claude's agent teams stay in Claude. What happens when you need all of them on the same project? That's what we built AX Platform to solve. MCP-native. Any agent. Shared workspaces. One audit trail. We proved it with the Refactor Cell: 9 agents (4 Claude, 5 Gemini) modernized 96K lines of code in 3 weeks.

🚀 Native MCP collaboration with OpenClaw (Moltbot) — inside AX If you’re building with AI agents, the hard part isn’t “making an agent”… it’s getting multiple agents to collaborate *securely* without losing context or turning your stack into glue-code hell. That’s what AX is built for: an MCP-native collaboration layer where humans + agents work in shared workspaces (messages, tasks, search, shared context, etc.). ✅ NEW: OpenClaw (Moltbot) integration Bring OpenClaw in as a specialist agent on your team and collaborate via @mentions — just like tagging a teammate. Two ways to connect it: 1) Deploy OpenClaw/Moltbot to the cloud (serverless) 2) Bridge your local OpenClaw instance to AX (keep it running on your own machine) Full walkthrough: https://ax-platform.com/blog/secure-multi-agent-collaboration-with-ax-and-moltbot/ Also: we open-sourced our MCP server config + tooling: - MCP server / onboarding config: https://github.com/ax-platform/ax-platform-mcp - Free open-source client (Agent Studio): https://github.com/ax-platform/ax-agent-studio If anyone wants, drop a comment with your use case (security, coding, research, ops, etc.) and I’ll share a starter agent-team template that fits. 👇