Quick Start
Get a MUR agent running and talking to you in about five minutes.
Install
See Installation for every route (Homebrew, install script, cargo, MUR Hub). The short one:
brew install mur-run/tap/mur
Initialize
mur init
An interactive wizard that:
- explains the recommended configuration (cloud LLM + local embedding)
- lets you choose an LLM provider (Anthropic, OpenAI, Gemini, OpenRouter, or Ollama)
- lets you choose an embedding model (Ollama or cloud)
- installs hooks for the AI tools it detects (Claude Code, Gemini CLI, …)
To install hooks without the prompts: mur init --hooks.
Create your first agent
mur agent create coach --model llama3.2:3b
The default provider is Ollama, so this needs no API key. To use a registered cloud model instead, pass its alias — see Model Registry.
Optionally run it as a real background service (launchd on macOS, systemd on Linux), so it survives logout and restarts:
mur agent install-service coach
Talk to it
murmur coach
murmur is the quick form of mur agent cli. You get a streaming TUI chat: tokens arrive live, tool calls show as step cards, and anything risky pauses for your approval. --resume continues the last conversation, and several names at once (murmur dev qa ops) open one tiled pane per agent.
Inside the chat, / opens a completion menu of every slash command and the agent's own skills; !cargo test runs on your machine and hands the output to the agent as your message. See Agent CLI for the full key map and the slash commands.
Teach the tools you already use
mur sync
This writes your knowledge into the configuration of every enabled AI tool, so Claude Code, Gemini CLI and the rest start each session already knowing your preferences. See Cross-CLI Sync.
Check on things
mur agent list # every agent and whether it is running
mur agent status coach
mur stats # statistics and effectiveness
mur dashboard # terminal dashboard
mur doctor # setup and configuration health
Stopping and removing
mur agent stop coach # unloads the service first, so the supervisor cannot respawn it
mur agent remove coach # unregisters it; --purge also deletes its data
What's next?
- Configuration — providers, embedding, and settings
- Agent Lifecycle — start, stop, services, and what changes once it runs as one
- Skills — the knowledge objects agents actually use
- Fleets — squads of agents working one goal
- Command Reference — all 34 commands and where each is documented