At Microsoft Build 2026, Microsoft launched the MXC SDK to enable OS-level containment for developer agents, addressing critical security and vulnerability concerns when running code generated by AI models. The buyer: DevOps leads and Engineering Directors who need sandboxed execution spaces to safely test and run AI-generated code.
active
AI Developer Tools
The tools agents and developers use to build — Claude Code, OpenClaw, MCP protocol, Codex CLI, and the emerging AI-native dev stack.
Last updated: Jun 3, 2026 Our take
We use these tools daily to build real products. Not reviews — field reports.
AI developer tools are reshaping how software gets built. We track the tools our agents use in production daily — not benchmarks, real work.
Tools We Use Daily
- Claude Code — primary builder tool for Kasra, Mumega agents
- OpenClaw — runs Athena, Sol, Worker, Dandan on various models
- MCP Protocol — the bus standard connecting all 17 agents
- Codex CLI — Codex agent’s primary interface
What We’re Watching
- Claude Code hooks and skills ecosystem
- MCP adoption across the industry
- Model context windows expanding (1M+ tokens)
- Agent-to-agent tool sharing
News & changes
Key Voices
Anthropic Claude Code, MCP creators github
OpenAI Codex CLI github
Sources
ART
Claude Code Documentation The CLI tool that powers our builder agents
ART Model Context Protocol The standard our entire bus runs on
GH Microsoft Execution Containers (MXC) SDK TypeScript SDK and schema to enforce OS-level boundaries (file, network, clipboard) for autonomous agents running in sandboxed execution environments.