Open Executive is a GitHub-hosted AI system that mimics a company's executive team through eight specialist agents — a Chief Strategy Officer, CFO, CHRO, General Counsel, COO, CMO, CPO, and Board Communications Director. All responses come from a single coherent voice (Claude Sonnet 4.6) that routes questions to the right specialists in parallel, pulling context from two sources: built-in MBA-level knowledge stored in ChromaDB and your company's own documents you upload. The system maintains episodic memory across sessions using SQLite, so it remembers past decisions and can proactively surface follow-ups via a built-in scheduler. It also supports prompt caching to hit 85% cache hit rates after a few turns, cutting API costs. The setup is straightforward for developers. You clone the repo, add your Anthropic API key to a .env file, and run `make dev` to spin up a FastAPI backend on port 8000 and a Next.js web UI on 3000. First boot takes a few minutes because it downloads a ~90 MB embedding model and pulls heavy ML dependencies (ChromaDB, PyTorch, sentence-transformers). After that it's fast. The system is single-instance only — the scheduler uses database locks to prevent double-firing, so don't try to horizontally scale without gating that first. You can integrate it with Discord, Slack, email, Telegram, and Google Chat through environment variables; the Discord bot, for example, requires you to create an app, enable the Message Content intent, and set three env vars. The onboarding wizard walks you through company basics (name, industry, stage, team size), business model, competitive landscape, strategic priorities, and culture. You can also upload documents like pitch decks or financial models via the web UI or CLI. The system is Apache 2.0 licensed and all configuration lives in that root .env file, which both the API and UI read at runtime. The architecture is modular — separate packages for the core orchestrator, eight agents, knowledge/RAG pipeline, memory, scheduler, and integrations — with detailed design docs in the repo.