1
Configure Agents
Define your council members with specific perspectives (optimist, skeptic, analyst, etc.)
2
Generate Files
Create agent prompts, shared_reasoning.md, and claude.md orchestration config
3
Invoke Council
Use trigger phrase to launch all agents in parallel with fresh context windows
4
Review & Decide
Get synthesized report with all perspectives documented in shared reasoning
Council Configuration
💡 Each agent runs in its own context window (~50k tokens), preventing the quality degradation that occurs at 40-50% context usage in single sessions.
Agent Builder
Generated Configuration
Click "Generate All Files" to create your AI Council configuration...
⚠️ Critical: Always test in a FRESH session after setup. Testing in the same session where you built the config will work due to in-context learning, but may fail in new sessions if config is incomplete.