Your repo, agent-ready in three commands.

What just happened

• leji start launched a detected agent host from the repository root, with one instruction: read the boot profile first, so its opening context is the team's rather than whatever it inferred from the tree. That is an instruction, not a check; leji points the agent at the file and never verifies that it read it. With no host detected, or in a scripted run, nothing launches and the command prints the line to run instead.
• Before it launched, leji start printed a Setup block for this clone: whether the Leji CLI this repository declares resolves here at the version the layer's spec line needs, whether the MCP server is registered for the host it selected, whether a shared .mcp.json is committed, and whether the pre-commit hook is installed. Each gap is reported with the command that closes it, and none of them blocks entry.
• Checking the layer itself is separate: leji validate for schemas and lint rules, leji conformance for the level it claims. Both run from the package you just installed, on your own machine, with no service in the loop.
• The tooling opens no connection of its own, and there is no account, no registry and no hosted service behind it. What reaches the network, you invoke yourself: the federation fetches contact the repository you named, and the dependency step in leji init and leji adopt runs your own package manager on your explicit yes. The agent leji launches is the one you already run, with its own network behavior.