what is mega-brain fr?

Mega Brain is a knowledge management system that turns raw material you own, such as videos, PDFs, and transcripts, into something an AI can reason with and cite. It runs inside Claude Code, Anthropic's command line coding assistant, using custom commands like ingest, process jarvis, conclave, and jarvis briefing. Everything is local first, meaning your material stays on your own computer's filesystem rather than living in someone else's database. The system works in three layers. First, an ingestion pipeline takes whatever you feed it, transcribes it if needed, breaks it into pieces, and pulls out what the project calls the cognitive DNA: things like the source's philosophies, mental models, heuristics, frameworks, values, and recurring patterns. Second, that extracted knowledge is indexed for hybrid search, meaning you can search by meaning as well as by keyword, and every answer points back to where it came from. Third, and described as the standout feature, is the Conclave: a panel of AI agents, including a methodological critic, a devil's advocate, and a synthesizer, that debate a strategic question you ask using only the knowledge you fed into the system, and nothing else. Setting it up requires Claude Code itself, Node.js 18 or newer, Python 3.10 or newer, and Git, plus API keys from external AI providers. An Anthropic key powers the core reasoning, an OpenAI key handles audio transcription and search embeddings, and a Gemini or Google key is needed if you want to ingest video. Optional keys add voice output, speech to text, and speaker-labeled transcription. The project is distributed as a private, invite only repository, and your API keys are kept in a local file that is never committed to the repository. This tool is built for people who want an AI assistant grounded strictly in material they trust, rather than one that answers from general internet knowledge, useful for founders, consultants, or researchers making decisions based on a specific body of expert content they have collected.