The AI Conclave turns your hardest decisions into governed deliberations — multiple AI minds at the table, preserved dissent, and a human-ratified audit trail. Because you need more than an answer. You need the reasoning trail.
"Most AI tools give you an answer. The AI Conclave gives you the reasoning trail — every voice, every disagreement, every rejected alternative — so you can make decisions you can actually stand behind."
What it does
Built for builders, writers, researchers, and technical creators who need more than a chatbot — they need governed deliberation with an audit trail.
Conclave — equal voices, full-mesh visibility. Consult — one primary, the rest as critics. Resolve — one agent drives, others advise. Pick the shape that fits your decision.
Disagreements are first-class records in the Conclave. The minority view is logged, exported, and permanently visible — not silently smoothed over by a false consensus.
The council deliberates. You ratify. Your decision is recorded as the authoritative final word, separate from the AI output. AI advises. You decide. Always.
Every exchange is logged to a local SQLite database. Export any task as PDF, Word, Markdown, or plain text. Decision records are tier-protected — they're never auto-deleted.
Runs entirely on 127.0.0.1. Your deliberations never pass through a Conclave server. API calls go directly to each provider — no middleman, no data escrow.
Attach your codebase. Agents receive a read-only snapshot — never the live files. They reason on your actual code, not a description of it. Your project is never modified.
A resolved task seeds the next one. The council carries forward your decision and the final answer as context. You build on what you've established — no re-explaining your starting point.
Live model pricing, daily spend charts, and per-agent token counts — built into the dashboard. Mix subscription CLIs with pay-per-token OpenRouter seats. A typical conclave turn costs $0.001–$0.02.
The Council
The Conclave isn't a chatbot wrapper. It seats AI participants from different labs — trained on different data, shaped by different post-training cultures. When they agree, you can trust it. When they disagree, that's the most valuable thing the session produces.
How it works
Choose your agents, your deliberation mode, and optionally attach your codebase or a URL. Permissions are explicit — agents only do what you explicitly allow, per task.
Agents exchange positions round by round. You watch it happen in real time on the dashboard — every turn, every disagreement, every shift in reasoning is visible as it happens.
Record your decision in your own words — what you chose, why, what you rejected. The record is stored permanently alongside the full transcript and exports as PDF, Word, or Markdown.
The Conclave Charter
The AI Conclave runs under a ratified charter — binding rules embedded into every agent prompt at runtime. Not a setting. Not a toggle. Structural.
Agents cannot write files, run commands, or install packages unless you explicitly grant each permission — per task, never globally. Default-deny, not default-allow.
Agents reason about your code. They don't run it. The deliberation is always about a stable, read-only snapshot. No surprises, no side-effects inside a session.
New features only land when the operational foundation is solid. No half-built capabilities, no silent failure modes. The charter is enforced in code, not in promises.
Ratified decisions are never deleted or edited — only superseded, with the original record preserved as audit history. The past is permanent. That's the point.
Get started
Runs locally on your machine. Bring your own subscriptions. First conclave in under five minutes.