PAL MCP (Collaborative Intelligence)
| Tool | When to Use | Purpose |
|------|-------------|---------|
| mcp__pal__consensus | Conflicting priorities | Multi-model resolution of trade-offs |
| mcp__pal__chat | Brainstorming | Collaborative idea exploration with external model |
| mcp__pal__thinkdeep | Complex problems | Multi-stage deep analysis |
| mcp__pal__planner | Solution design | Sequential planning with branching |
| mcp__pal__challenge | Validate ideas | Force critical thinking on proposed solutions |
PAL Usage Patterns
```bash
# Consensus on conflicting priorities
mcp__pal__consensus(
models=[
{"model": "gpt-5.2", "stance": "for", "stance_prompt": "Prioritize user experience"},
{"model": "gemini-3-pro", "stance": "against", "stance_prompt": "Prioritize technical simplicity"},
{"model": "deepseek", "stance": "neutral"}
],
step="Evaluate: Should we use real-time sync or eventual consistency?"
)
# Deep exploration of complex idea
mcp__pal__thinkdeep(
step="Exploring AI-powered analytics dashboard concept",
hypothesis="Users need predictive insights, not just historical data",
confidence="medium",
focus_areas=["user_needs", "technical_feasibility", "market_fit"]
)
# Collaborative brainstorming
mcp__pal__chat(
prompt="Help me explore innovative approaches for real-time collaboration in document editing",
model="gpt-5.2",
thinking_mode="high"
)
# Challenge assumptions
mcp__pal__challenge(
prompt="We assume users want AI-generated summaries. Is this assumption valid?"
)
# Plan solution architecture
mcp__pal__planner(
step="Planning architecture for real-time notification system",
step_number=1,
total_steps=4,
is_branch_point=True,
branch_id="websocket-approach"
)
```
Rube MCP (Research & Persistence)
| Tool | When to Use | Purpose |
|------|-------------|---------|
| mcp__rube__RUBE_SEARCH_TOOLS | Market research | Find web search, competitor analysis tools |
| mcp__rube__RUBE_MULTI_EXECUTE_TOOL | Documentation | Save ideas to Notion, share in Slack |
| mcp__rube__RUBE_CREATE_UPDATE_RECIPE | Workflows | Save brainstorming processes |
| mcp__rube__RUBE_REMOTE_WORKBENCH | Data analysis | Analyze market data, user research |
Rube Usage Patterns
```bash
# Research market and competitors
mcp__rube__RUBE_SEARCH_TOOLS(queries=[
{"use_case": "web search", "known_fields": "query:AI analytics dashboard competitors 2025"}
])
# Document brainstorming session
mcp__rube__RUBE_MULTI_EXECUTE_TOOL(tools=[
{"tool_slug": "NOTION_CREATE_PAGE", "arguments": {
"title": "Brainstorm: AI Analytics Dashboard",
"content": "## Key Ideas\n- Predictive insights\n- Natural language queries\n\n## Decisions\n- Real-time sync chosen over eventual consistency"
}},
{"tool_slug": "SLACK_SEND_MESSAGE", "arguments": {
"channel": "#product",
"text": "New brainstorm session documented: AI Analytics Dashboard"
}}
])
# Create user research tasks
mcp__rube__RUBE_MULTI_EXECUTE_TOOL(tools=[
{"tool_slug": "JIRA_CREATE_ISSUE", "arguments": {
"project": "PROD",
"summary": "User research: AI analytics preferences",
"issue_type": "Task",
"description": "Interview 10 users about analytics needs"
}},
{"tool_slug": "ASANA_CREATE_TASK", "arguments": {
"name": "Competitor analysis: analytics dashboards",
"project": "Research"
}}
])
# Analyze existing user feedback
mcp__rube__RUBE_REMOTE_WORKBENCH(
thought="Analyze user feedback data for patterns",
code_to_execute='''
import json
# Load user feedback from file
feedback_data = json.load(open("/tmp/user_feedback.json"))
# Analyze with LLM
analysis, error = invoke_llm(f"Analyze this user feedback for analytics feature requests: {feedback_data[:5000]}")
output = {"analysis": analysis, "feedback_count": len(feedback_data)}
output
'''
)
```
|