Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ruvnet/ruflo/llms.txt

Use this file to discover all available pages before exploring further.

ruflo agent is the primary interface for controlling the lifecycle of individual AI agents. Use it to spawn specialized workers, inspect their state and metrics, drain logs in real time, and manage the auto-scaling pool. Agents are the execution units that make up a swarm — each one is typed, named, and tracked through the MCP layer.

Synopsis

ruflo agent <subcommand> [options]

Subcommands

SubcommandAliasDescription
spawnSpawn a new agent of a given type
listlsList all active (or all) agents
statusShow detailed status and metrics for one agent
stopkillGracefully (or forcefully) stop an agent
metricsShow aggregate performance metrics
poolView and configure the agent auto-scaling pool
healthHealth dashboard for all agents or a single agent
logsStream or tail structured logs for an agent

Agent types

TypeCapabilities
coderCode generation, refactoring, debugging, testing
researcherWeb search, data analysis, summarization, citation
testerUnit testing, integration testing, coverage analysis, automation
reviewerCode review, security audit, quality check, documentation
architectSystem design, pattern analysis, scalability, documentation
coordinatorTask orchestration, agent management, workflow control
analystPerformance analysis, optimization, benchmarking
optimizerPerformance optimization, bottleneck analysis
security-architectThreat modeling, security patterns, compliance, audit
security-auditorCVE remediation, security testing
memory-specialistAgentDB unification, vector search, caching, optimization
swarm-specialistUnified coordination engine
performance-engineerBenchmarking, profiling, optimization, monitoring
core-architectDomain-driven design restructure
test-architectTDD London School methodology

spawn

Create and register a new agent. The agent is persisted through the MCP layer and reflected in .claude-flow/metrics/swarm-activity.json so the statusline picks it up immediately.
ruflo agent spawn --type <type> [options]
--type / -t
string
required
Agent type to spawn. See the agent types table above for valid values.
--name / -n
string
Human-readable identifier for this agent. Auto-generated as <type>-<timestamp> if omitted.
--model / -m
string
LLM model override for this agent.
--provider / -p
string
default:"anthropic"
Provider to use: anthropic, openrouter, or ollama.
--task
string
Initial task description passed to the agent on start.
--timeout
number
default:"300"
Agent timeout in seconds before automatic termination.
--auto-tools
boolean
default:"true"
Allow the agent to use tools autonomously.

list

List agents visible to the MCP layer.
ruflo agent list [options]
--all / -a
boolean
default:"false"
Include terminated and idle agents in addition to active ones.
--type / -t
string
Filter output to agents of this type only.
--status / -s
string
Filter by status: active, idle, or terminated.

status

Show the full status record for a single agent, including its live metrics.
ruflo agent status <agentId> [--id <id>]
--id
string
Agent ID (can also be passed as the first positional argument).
The output table includes: agent type, status, creation time, last activity, tasks completed, tasks in progress, tasks failed, average execution time, and uptime.

stop

Gracefully shut down an agent (saves state and releases resources) or force-terminate it immediately.
ruflo agent stop <agentId> [options]
--force / -f
boolean
default:"false"
Skip the graceful shutdown sequence and send an immediate SIGKILL equivalent.
--timeout
number
default:"30"
Seconds to wait for graceful shutdown before giving up.
Stopping an agent updates .claude-flow/metrics/swarm-activity.json so the statusline agent count decrements immediately.

metrics

Show aggregate metrics sourced from .swarm/agents/*.json and .swarm/swarm-activity.json.
ruflo agent metrics [agentId] [--period <period>]
--period / -p
string
default:"24h"
Time window for metrics aggregation. Valid values: 1h, 24h, 7d, 30d.
Output includes: total agents, active agents, tasks completed, success rate, memory vector count, and a per-type breakdown.

pool

View and configure the agent auto-scaling pool. Agents in the pool are pre-warmed and assigned tasks without the cold-start penalty of agent spawn.
ruflo agent pool [options]
--size / -s
number
Set a fixed pool size immediately.
--min
number
default:"1"
Minimum agents to keep alive.
--max
number
default:"10"
Maximum agents allowed in the pool.
--auto-scale / -a
boolean
default:"true"
Enable demand-based auto-scaling within the --min/--max range.

health

Display a health dashboard across all agents (or a single agent with --id).
ruflo agent health [options]
--id / -i
string
Restrict output to a single agent.
--detailed / -d
boolean
default:"false"
Show per-agent uptime, latency (avg + p99), task breakdown, and error log.
--watch / -w
boolean
default:"false"
Refresh the dashboard every 5 seconds.
The overall summary shows counts of healthy, degraded, and unhealthy agents alongside average CPU and memory utilization.

logs

Tail or stream structured JSON logs emitted by an agent.
ruflo agent logs --id <agentId> [options]
--id / -i
string
required
Agent ID whose logs to retrieve.
--tail / -n
number
default:"50"
Number of recent log entries to show.
--level / -l
string
default:"info"
Minimum log level to display: debug, info, warn, or error.
--follow / -f
boolean
default:"false"
Stream new log entries as they arrive.
--since
string
Show only entries since this duration ago, e.g. 1h or 30m.

Examples

# Spawn a coder agent with an auto-generated name
ruflo agent spawn -t coder

# Spawn a researcher with a specific name and task
ruflo agent spawn -t researcher -n scout-1 --task "Research React 19 new features"

# Spawn an architect using a specific model
ruflo agent spawn -t architect --model claude-opus-4-5-20251101
Use ruflo hive-mind spawn "<objective>" as a higher-level alternative to agent spawn when you want Ruflo to automatically select and coordinate multiple agent types for a complex goal.

Build docs developers (and LLMs) love