Work with coding agents

Give an agent its own development environment on your server.

An isolated place to work

A workspace is a virtual development machine with its own filesystem and tools. Run the agent there instead of directly in your laptop’s normal working environment. Supply the credentials and permissions that agent needs deliberately.

Isolation defines the working environment; it does not make arbitrary agent commands harmless. Review credentials, source changes and any access you grant.

Keep a session running

Connect to the Ready workspace, then start a persistent terminal session before launching a compatible agent CLI:

tmux new -s agent

Launch your configured coding agent inside that session. Detach with Ctrl+B, then D. After reconnecting to the workspace, reattach with:

tmux attach -t agent

Disconnecting SSH leaves that session on the node. Stopping the workspace or powering off the node stops its processes. Agent CLI installation, authentication and provider requirements remain specific to that agent.

Connect through MCP

Readylab also provides constrained MCP tools for project, workspace and deployment operations. An MCP host launches readylab-mcp over stdio as the daemon owner. For a remote node, the command can run through an already verified SSH alias:

ssh -T readylab-node readylab-mcp

The daemon must already be running and the MCP process must have access to its socket. Configure the MCP host’s command and arguments for your client; the alias and executable path depend on your installation.

Product operations, bounded access

The MCP adapter exposes product operations, readiness and connection metadata. It does not expose arbitrary shell execution, file read/write, raw runtime control or service credentials. Agents use their own configured identity to connect to a workspace.