Claw Multi Agent

Same round → parallel execution:

sessions_spawn(task="Search LangChain...") ──┐
sessions_spawn(task="Search CrewAI...")    ──┤→ all run simultaneously
sessions_spawn(task="Search AutoGen...")   ──┘
sessions_spawn(task="Search LangGraph...") ─┘

↓  (all finish, main agent receives all 4 results)

Main agent consolidates → writes full report

# Parallel research: spawn all 4 in the same round → they run simultaneously
sessions_spawn({
    "task": "[CONTEXT] Comparing AI agent frameworks for a tech team report.\n\n[YOUR TASK] Search LangChain: architecture, pros/cons, GitHub stars, latest version. Return 5 bullet points ≤100 words each. Do NOT write a full report.",
    "label": "🔍 researcher-langchain [model: default]"
})
sessions_spawn({
    "task": "[CONTEXT] Same report.\n\n[YOUR TASK] Search CrewAI: architecture, pros/cons, GitHub stars, latest version. Return 5 bullet points ≤100 words each.",
    "label": "🔍 researcher-crewai [model: default]"
})
sessions_spawn({
    "task": "[CONTEXT] Same report.\n\n[YOUR TASK] Search AutoGen: architecture, pros/cons, GitHub stars, latest version. Return 5 bullet points ≤100 words each.",
    "label": "🔍 researcher-autogen [model: default]"
})
sessions_spawn({
    "task": "[CONTEXT] Same report.\n\n[YOUR TASK] Search LangGraph: architecture, pros/cons, GitHub stars, latest version. Return 5 bullet points ≤100 words each.",
    "label": "🔍 researcher-langgraph [model: default]"
})
# All 4 run in parallel → when all return, main agent consolidates and writes report

# Phase 1: parallel research (spawn all at once)
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search LangChain. 5 bullets ≤100 words.", "label": "🔍 researcher-langchain"})
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search CrewAI. 5 bullets ≤100 words.", "label": "🔍 researcher-crewai"})
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search AutoGen. 5 bullets ≤100 words.", "label": "🔍 researcher-autogen"})

# Phase 2: after all 3 return → main agent writes report (sequential, depends on research)
# (main agent does this directly, no need to spawn a writer)

cd ~/.openclaw/skills/claw-multi-agent

# Parallel: multiple agents tackle different angles simultaneously
python run.py --mode parallel \
  --agents "fast:🔍 researcher:summarize the pros of microservice architecture" \
           "fast:🔍 researcher:summarize the cons of microservice architecture" \
           "fast:🔍 researcher:list real-world companies using microservices and outcomes" \
           "smart:📊 analyst:compare microservices vs monolith for a 10-person startup" \
  --aggregation synthesize

# Sequential: chain agents, each builds on the previous output
python run.py --mode sequential \
  --agents "fast:📋 planner:break down how to build a REST API in Python" \
           "smart:💻 coder:implement the API based on the plan above" \
           "fast:🔎 reviewer:review the code for bugs and security issues" \
  --aggregation last

# Auto-route: router classifies task and picks tiers automatically
python run.py --auto-route --task "write a technical blog post about GRPO vs PPO"

# Dry-run: preview the plan without executing
python run.py --dry-run \
  --agents "fast:researcher:research X" "smart:writer:write report"

Phase 1 (Orchestrator — with tools, parallel):
  sessions_spawn(search topic A) ──┐
  sessions_spawn(search topic B) ──┤ → all run simultaneously
  sessions_spawn(search topic C) ──┘
  ↓ research summaries collected

Phase 2 (Pipeline — pure text, parallel):
  openclaw agent (writer style 1) ──┐
  openclaw agent (writer style 2) ──┤ → all run simultaneously
  openclaw agent (writer style 3) ──┘
  ↓ 3 draft versions returned

Main agent: compare drafts → pick best or synthesize

# Auto: router detects hybrid intent and runs both phases
python run.py --mode hybrid --task "调研主流AI框架，给我3个不同风格的对比报告" --num-drafts 3

# Auto-mode: let router decide the mode automatically
python run.py --auto-mode --task "搜索竞品资料后写几个版本的分析"

# Phase 1: parallel research (spawn all at once)
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search LangChain. 5 bullets.", "label": "🔍 research-langchain"})
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search CrewAI. 5 bullets.", "label": "🔍 research-crewai"})
sessions_spawn({"task": "[CONTEXT] ...\n\n[TASK] Search AutoGen. 5 bullets.", "label": "🔍 research-autogen"})

# After all 3 return → Phase 2: main agent writes 3 draft versions itself
# (or spawn 3 pipeline agents with research as context)

python scripts/router.py classify "write a Python web scraper"
# → Tier: CODE  (routes to smart model)

python scripts/router.py classify "research the latest LLM papers"
# → Tier: RESEARCH  (routes to fast model)

python scripts/router.py spawn --json --multi "research X and write a report"
# → splits into 2 tasks: RESEARCH + CREATIVE

[CONTEXT] User is comparing AI agent frameworks for a team report. Audience: engineers.

[YOUR TASK] Search LangChain pros and cons. Return 5 bullet points ≤100 words each.

[PRIOR FINDINGS]
- LangChain: richest ecosystem, steep curve
- CrewAI: clean role separation...

[YOUR TASK] Based on above, search AutoGen. Return 3 unique points not covered above.

[CONTEXT FILE] Read /workspace/research/context.md for full background.

[YOUR TASK] Search latest Test-Time Compute Scaling advances. Return 3 summaries.

BG = "Researching RL post-training for ML engineers. Topics: GRPO/DAPO/PPO, veRL."

sessions_spawn({"task": f"[CONTEXT] {BG}\n\n[TASK] Search GRPO vs PPO benchmarks. 5 bullets ≤100 words.", "label": "🔍 researcher-grpo [model: default]"})
sessions_spawn({"task": f"[CONTEXT] {BG}\n\n[TASK] Search DAPO design. 5 bullets ≤100 words.", "label": "🔍 researcher-dapo [model: default]"})
sessions_spawn({"task": f"[CONTEXT] {BG}\n\n[TASK] Search veRL architecture. 5 bullets ≤100 words.", "label": "🔍 researcher-verl [model: default]"})

## 📊 Execution Summary

Mode: 🎯 Orchestrator Mode (sessions_spawn, with tools)

| Agent | Role | Model | Time | Status |
|-------|------|-------|------|--------|
| 🔍 researcher-langchain | Researcher | default | 22s | ✅ |
| 🔍 researcher-crewai    | Researcher | default | 19s | ✅ |
| 🔍 researcher-autogen   | Researcher | default | 24s | ✅ |
| 🔍 researcher-langgraph | Researcher | default | 21s | ✅ |
| ✍️ main (consolidate)   | Writer     | default | 38s | ✅ |

Agents spawned: 4  |  Parallel time: ~24s  |  Serial equivalent: ~86s  |  Saved: ~62s (72%)

python run.py
  --mode parallel|sequential
  --agents "tier_or_model:🎭role:task description"   # repeatable, any number
  --aggregation synthesize|compare|concatenate|last
  --timeout 300
  --dry-run          # preview without executing
  --auto-route       # router picks tiers automatically
  --list-models      # show current model config