Synopsis
Prerequisites
The command checks three things before starting the session:claudeis on yourPATH(install withnpm install -g @anthropic-ai/claude-code).- The murmur MCP server is configured in Claude Code (run
claude mcp add murmur -- murmur mcpif not). - Claude Code is authenticated (run
claude authif not).
Why murmur director passes --dangerously-load-development-channels
Under the hood,
murmur director launches Claude Code with --dangerously-load-development-channels server:murmur. The flag opts the session into Claude Code’s channel transport, which Murmur uses to stream real-time events (PR comments, CI results, child lifecycle, follow-ups) from the control plane into your conversation as <channel> tags. Without it the director can’t receive event pushes, and the coordinator loop across multiple agents stops working.The dangerously- prefix is Claude Code’s wording for an opt-in developer feature, not a security warning specific to Murmur. The same flag is also used by murmur init and by every interactive agent VM that Murmur spawns, so all sessions can post to and receive from server:murmur.When Claude Code prompts you to confirm channels, say yes and opt in. Murmur can’t deliver event pushes without it.Arguments
Examples
Start a director session with the default persona
murmur-director persona, then opens an interactive Claude Code session with the murmur MCP server attached. You land in a conversational interface where you can give instructions, spawn agents, check status, and coordinate work.
Start a session with a custom persona
code-reviewer persona from the catalog and starts the session with that persona’s prompt.
Errors
Related
- Director, concept overview of what a director is and the behavioral contract it follows
- Agent Personas, what personas are and how they shape agent behavior
- Catalog, where personas and other resources live
- Agents, concept overview
murmur status, check agent status during a director session- Multi-agent orchestration, guide to coordinating agents as a director