AI Pulse · Under the Radar

Three separate projects landed this week, all solving the same problem: if you’re running multiple AI agent tools, you need one control panel to manage them instead of jumping between tabs and terminals.

The first is cc-switch, a GitHub repo that hit 121,000 stars and is adding 341 per day. It lets you switch between eight different coding-agent providers from a single interface and handles MCP server connections and skill routing in one place. The second is BossConsole, a JVM-based operator’s console built for enterprises. It runs Claude Code, Codex, and Gemini agents with browser automation, terminal access, secrets management, and 100 MCP tools under one roof. It launched with 39 stars per day, which is quiet but steady for enterprise tooling. The third is openworker, quietly open-sourced by Andrew Ng’s team. It already has 4,700 stars and 620 forks. It’s a local-first “AI coworker” with 25-plus integrations and approval gates before any consequential action runs. Chinese AI communities were dissecting it days before Western coverage picked it up.

Why this matters

If you’re testing or deploying agents, you’re probably running more than one. Maybe Claude for code, Gemini for research, something else for browser tasks. Each tool has its own interface, its own quirks, its own way of handling credentials and tool access. That friction compounds fast. These three projects, from different corners of the world, all converged on the same answer: one control surface over many harnesses. That’s not a coincidence. It’s a pattern worth watching, and it’s exactly the kind of orchestration layer we build into the Omni Command Centre when businesses need to coordinate multiple agents without duct tape and prayer. If you’re juggling tools today, expect this category to move quickly.