Coder

If your CIO asked what AI ran last week, and what it cost, could you answer in under a minute? Most platform teams can't, and it's not a tooling gap, it's an ownership gap. Provisioning, policy, and permissions for developers is a job you already do. The AI Operating Layer is that same job, widened from developers to everyone running AI. No reorg required.

The person running AI at your company isn't only Engineering. It's Finance building a model. It's HR running an agent. Nobody logged any of it, and nobody's in trouble for it, that's just where adoption starts now. The AI Operating Layer is how platform teams extend the same safety net they already built for developers to everyone else.

In July, two frontier AI labs lost containment inside environments built specifically to hold it. Not a bad actor. Not bad judgment. The environment gave way. If it can happen there, the question isn't whether your AI is safe today. It's what's actually holding it. The AI Operating Layer is the case for governing where AI runs, not which tools people use.

Banning tools doesn't work. People route around policy the moment it slows them down. Control that depends on good behavior isn't control. The AI Operating Layer contains data, scopes access, and governs models underneath whatever tool your team already picked: Claude Code, Cursor, ChatGPT, all of it. You already own the closest thing to this layer. This is the mandate to build it.

Claude Code is Now Available with Coder Agent Relay! With AI safety in the headlines, enterprises keep asking “where can agents actually safely run?” The answer now: on the customer's own infrastructure. Customers can use Anthropic's agents, and Anthropic still runs the Claude Code agent loop. Agent Relay connects it to Coder workspaces on the customer's cloud, VPC, or on-prem environment, where commands, file operations, and other tool calls execute. This is a different way to run AI. Banks, government agencies and global enterprises can now say yes to Claude Code without giving up the security posture they've built. It follows our Cursor/SpaceXAI launch two weeks ago and brings customers another step closer to any agent running safely inside the enterprise. Your org can now use Claude Code with Coder, read more at https://lnkd.in/dENvWH_e

Banning tools doesn't work. People route around policy the moment it slows them down. Control that depends on good behavior isn't control. The AI Operating Layer contains data, scopes access, and governs models underneath whatever tool your team already picked: Claude Code, Cursor, ChatGPT, all of it. You already own the closest thing to this layer. This is the mandate to build it.