A multi-agent system that debates your ideas, stress-tests assumptions with a built-in adversary, and surfaces a verdict you can act on.
Runs locally. Uses subscription CLI tools — Claude, Gemini, Codex — so there are no per-query API costs. Works in Discord or as a simple local markdown file.
No API key. No install. Runs on Groq free tier.
A single model reflects your framing back at you. It agrees with the premise, fills in your blind spots with optimism, and rarely surfaces the strongest objection.
Running multiple models via API for every question gets expensive fast. Most people end up asking one model anyway — or paying for subscriptions they barely use.
Switching between Claude, Gemini, and ChatGPT for different questions is friction. Most people just stick with one — even when another would be better.
Agent Advisory Board routes your question to the right advisors automatically, forces one to argue against the consensus, then synthesizes — so the output already reflects the best case and the worst case. You get a verdict, not a draft.
Every debate automatically assigns one agent as Red Team — an adversarial role that attacks the dominant view, not to be contrarian for its own sake, but to surface real weaknesses before you act.
One agent is told: attack the dominant view, find the worst-case scenario, refuse easy consensus. It's explicitly adversarial — stress-testing the position before you commit to it.
Claude synthesizes the full transcript — noting which Red Team objections were valid vs. weak. Points of agreement reached under adversarial pressure are flagged as stronger signal.
Type !board: topic and the system picks which advisors are relevant based on the topic. Or force specific ones: !board cfo cmo: topic
Advisors also work in debates: !debate cfo vs cmo: Should we raise prices?
!cfoUnit economics, burn rate, ROI timelines. Skeptical of optimistic projections.
!cmoWho's the customer and why do they buy? GTM sequencing, competitive position.
!ctoBuild vs buy, feasibility, what breaks at 10x scale.
!cooWho owns this? By when? What's blocking us?
!cpoAre we solving a real pain? For whom? How do we know?
!gcWhat's the legal exposure? What's missing from the contracts?
!uxHas anyone watched a real user try this? Where do they hesitate? What do they read first?
!grandmaWarm, practical, completely unimpressed by buzzwords. Will real people actually use this?
!teenagerBrutally honest Gen Z radar for cringe. Is this actually cool or are you trying too hard?
!neighborSeen every scheme fail for 30 years. What's the obvious way this goes wrong?
!internMaximum enthusiasm, zero cynicism. What if we just automated the whole thing?
!sharkWhat are your numbers? No patience for vanity metrics or TAM hallucinations.
ADVISORS_JSON in .env — any persona, any lens, any voice.Same engine, same agents, same debates. Pick whichever interface fits how you work.
No install. No API key. Open your browser and start a debate. Runs on Groq free tier — fast, free, and shareable.
Best for teams or if you want a persistent shared log. Bot joins your server, listens to a channel. Results appear as rich embeds with Approve / Reject / Ask Claude buttons.
No Discord account needed. Two files: inbox.md (you type here) and conversation.md (growing log). Open side-by-side in VS Code or Obsidian.
TRANSPORT=both in your .env to run Discord bot + local markdown simultaneously.Agent Advisory Board uses CLI tools that authenticate with your existing subscriptions. Configure only what you have — unconfigured agents are skipped automatically.
Anthropic's Claude via CLI. Uses your existing Max subscription — no per-query cost.
Google Gemini via CLI. Authenticates with your Google account. No API key needed.
Any local model (llama3.2, mistral, etc.) running on your machine. Free and private.
OpenAI Codex via desktop app. Requires an OpenAI subscription — no token charges.
Also supports: Ollama remote nodes (another machine on Tailscale), any llama.cpp server via BEAST_URL.
| Command | What it does |
|---|---|
What is X? | Auto-routed to best available agent |
!claude: prompt | Force Claude CLI |
!gemini: prompt | Force Gemini CLI |
!board: topic | Auto-select relevant advisors by topic |
!board cfo cmo: topic | Force specific advisors |
!board grandma shark: topic | Mix professional + fun advisors |
!debate: topic | 2-agent debate, default agents |
!debate cfo vs cmo: topic | Advisor debate — CFO vs CMO personas |
!debate claude vs gemini --red gemini: topic | Gemini plays Red Team |
!debate claude vs gemini --socratic cfo: topic | CFO plays Socratic Examiner |
!approve (file) / ✅ button (Discord) | Execute the proposed action |
!reject (file) / ❌ button (Discord) | Dismiss — no action |
!ask (file) / 🔍 button (Discord) | Ask Claude to elaborate |
Clone the repo, copy .env.example, set which agents you have. Takes 2 min.
Discord bot for shared team use. Local markdown file if you just want a fast personal setup.
One command. The coordinator starts polling. Bot comes online or inbox.md appears, ready.
Type in Discord or inbox.md. Agents respond, debate, synthesize. Click Approve or type !approve.
After cloning the repo, open Claude Code in the agent-factory/ directory and run the setup skill. Claude will ask which agents you have, which interface you want, walk through Discord bot creation step-by-step, set up your database, write your .env, and run the first test.
git clone https://github.com/rdcahalane/ai-skills.git cd ai-skills/agent-factory claude # opens Claude Code # then type: /setup-agent-factory