Documentation Index
Fetch the complete documentation index at: https://docs.murmur.dev/llms.txt
Use this file to discover all available pages before exploring further.
Once an agent is spawned, you need to check its status, send follow-ups, watch its progress, and eventually stop or kill it. This page covers all the agent management commands.
murmur status
Query the current state of an agent, including phase, VM info, progress, cost, PRs, and timeline events.
Flags
| Flag | Type | Description |
|---|
--service | bool | Show the underlying service details |
--children | bool | Include child agent statuses in the response |
--timeline | string | Comma-separated timeline event types to include |
--workspace | string | Override workspace |
Timeline event types
| Type | Description |
|---|
phase | Phase transitions (provisioning, running, idle, etc.) |
child | Child agent lifecycle events |
taskComplete | Task checklist item completions |
commit | Git commits made by the agent |
push | Git pushes |
checks | CI check results |
followUp | Follow-up messages delivered |
upload | File uploads |
# Basic status
murmur status fix-auth-bug
# Status with children and timeline
murmur status fix-auth-bug --children --timeline phase,commit,checks
# Service-level details
murmur status fix-auth-bug --service
Example output
Agent: fix-auth-bug
Phase: running
VM: vm-a1b2c3d4 (us-central1-a)
Model: claude-opus-4-6
Cost: $1.23
Progress:
[x] Identified the bug in session.go
[x] Implemented the fix
[ ] Running tests
PRs:
#142 - Fix session timeout handling (open)
murmur ls
List your running agents. By default, shows only your own active agents.
Flags
| Flag | Short | Type | Description |
|---|
-a | | bool | Include completed, failed, and killed agents |
-q | | bool | Quiet mode — print slugs only, one per line |
--json | | bool | Output as JSON array |
-A | | bool | Show all developers’ agents (admin) |
--developer | | string | Filter to a specific developer login, or * for all |
--workspace | -w | string | Override workspace |
# List your active agents
murmur ls
# Include completed/failed agents
murmur ls -a
# JSON output for scripting
murmur ls --json
# Slugs only (pipe to other commands)
murmur ls -q
# All developers' agents
murmur ls -A
# Specific developer
murmur ls --developer alice
Example output
SLUG PHASE VM COST AGE
fix-auth-bug running vm-a1b2c3d4 $1.23 12m
update-docs idle -- $0.45 1h
deploy-fix sleeping -- $2.10 3h
murmur wait
Block until an agent reaches a target phase. Useful in scripts and CI pipelines.
murmur wait <slug> [flags]
Flags
| Flag | Type | Description |
|---|
--phase | string | Phase to wait for (e.g., task-complete, idle, running). Defaults to any phase needing attention. |
--workspace | string | Override workspace |
# Wait for agent to complete
murmur wait fix-auth-bug --phase task-complete
# Wait for agent to reach idle (finished its turn)
murmur wait fix-auth-bug --phase idle
# Wait for any notable phase change (default)
murmur wait fix-auth-bug
murmur kill
Cancel an agent and its entire subtree of child agents.
murmur kill <slug> [flags]
Flags
| Flag | Type | Description |
|---|
--service | bool | Also terminate the underlying service |
--workspace | string | Override workspace |
# Kill an agent and all its children
murmur kill fix-auth-bug
# Kill with service cleanup
murmur kill fix-auth-bug --service
murmur kill is irreversible for the current session. The agent’s branch and any pushed code remain intact, but the conversation context is lost unless you --resurrect later.
murmur sleep / murmur wake
Sleep stops the VM while preserving the agent’s state. Wake restarts the agent on a fresh VM with its full context restored.
murmur sleep <slug>
murmur wake <slug>
# Put an agent to sleep (stops the VM, saves money)
murmur sleep fix-auth-bug
# Wake it up later (claims a new VM, restores context)
murmur wake fix-auth-bug
murmur session
Interact with a live agent’s Claude Code session.
session watch
Stream the agent’s session output to your terminal in real time:
murmur session watch <slug>
session send
Send a message to the agent’s active Claude session:
murmur session send <slug> "Also update the tests for the new behavior"
session interrupt
Interrupt the agent’s current Claude turn (equivalent to Ctrl+C):
murmur session interrupt <slug>
session stop
Stop the Claude Code process on the VM (the agent remains active):
murmur session stop <slug>
session send delivers a message directly to the active Claude session. For messages that should be queued and delivered between turns, use murmur queue add instead.
murmur queue
Manage the follow-up message queue for agents. Messages in the queue are delivered when the agent is ready (between Claude turns).
queue add
Send a follow-up message to a running agent:
murmur queue add <slug> "Also fix the session expiry logic"
murmur queue add <slug> "New requirements came in" \
--task "Add rate limiting" \
--task "Update API docs"
queue clear
Discard all queued follow-ups for an agent:
murmur queue clear <slug>
queue strategy
Change the dequeue strategy for an agent. This controls how many queued messages are delivered per turn:
murmur queue strategy <slug> <strategy>
| Strategy | Behavior |
|---|
all (default) | Drain all queued follow-ups at once |
one | Deliver one message per turn |
five | Deliver up to five messages per turn |
murmur queue strategy fix-auth-bug one
queue conflict-resolution
Configure how conflicting follow-ups are resolved:
murmur queue conflict-resolution <slug> <mode>
murmur each
Fanout: spawn one agent per line from stdin. The {} placeholder in the slug template and description template is replaced with each line.
murmur each [flags] <slug-template> <desc-template>
Flags
| Flag | Short | Type | Description |
|---|
--dry-run | | bool | Print what would be spawned without actually spawning |
--agent | | string | Agent persona for all spawned agents |
--model | | string | Model override for all spawned agents |
-d | | string | Delimiter for splitting lines (default: newline) |
--repo | | string | Repository override. Repeatable. |
--workspace | -w | string | Override workspace |
# Spawn one agent per module to fix tests
echo -e "auth\npayments\nnotifications" | murmur each "fix-{}" "Fix failing tests in the {} module"
# Dry run to preview
cat modules.txt | murmur each --dry-run "fix-{}" "Fix the {} module"
# With model override
cat packages.txt | murmur each --model claude-opus-4-6 "refactor-{}" "Refactor the {} package"
Example dry-run output
Would spawn: fix-auth "Fix failing tests in the auth module"
Would spawn: fix-payments "Fix failing tests in the payments module"
Would spawn: fix-notifications "Fix failing tests in the notifications module"
murmur rekey
Rotate credentials on a running agent and all its children. Re-encrypts and re-delivers your developer profile without restarting the VM.
# Rotate credentials after a token refresh
murmur rekey fix-auth-bug
murmur watch
Watch an agent’s events in real time. Similar to session watch but shows structured events (phase changes, commits, task completions) rather than raw session output.
murmur watch <slug> [flags]
Flags
| Flag | Type | Description |
|---|
--from | string | Start from a specific timestamp (RFC3339) |
--json | bool | Output events as JSON lines |
# Watch events in real time
murmur watch fix-auth-bug
# JSON output for processing
murmur watch fix-auth-bug --json
# Start from a specific time
murmur watch fix-auth-bug --from 2025-01-15T10:00:00Z
Example output
10:23:01 phase → running
10:23:45 commit abc1234 "Fix session timeout handling"
10:24:12 push branch fix-auth-bug → origin
10:24:30 checks CI started (3 checks)
10:25:15 checks CI passed (3/3)
10:25:20 phase → idle
