Devs Replaced by AI Answer Back With an Open-Source AI CEO

Open Executive, released by Sente Labs, is a tongue-in-cheek rejoinder to executives who cut engineering headcount in favor of AI: an open-source system that plays the role of a company’s entire C-suite. It presents a single executive persona to the user while fanning out, via tool calls, to eight specialist agents covering strategy, finance, HR, legal, operations, marketing, product, and board communications. The orchestration runs on Claude Sonnet with a Haiku pass for memory, and the multi-agent plumbing is deliberately hidden behind one coherent voice.

The technically interesting parts are in how it grounds and remembers. Each specialist pulls from two RAG layers in ChromaDB—git-tracked ‘MBA-level’ reference material plus the user’s own uploaded company documents—with retrieved context injected into the user turn rather than the cached system prompt, preserving a claimed 85% prompt-cache hit rate. Episodic memory extracts past decisions into SQLite so later sessions open with what was recommended before, and a built-in scheduler claims due jobs via UPDATE … RETURNING to avoid double-firing. That last detail carries a hard operational constraint: the API must run as a single instance, and the Fly.io configs pin max_machines_running to 1.

Beyond the satire, it’s a fairly complete reference implementation of a memory-backed, multi-agent assistant: FastAPI core, Next.js 15 UI, an onboarding wizard that builds a company profile, integrations for Slack, Discord, Telegram, Email, and Google Chat, plus eval scenarios with an LLM-as-judge runner. The framing is the joke, but the repo is a usable blueprint for anyone building retrieval-grounded agent orchestration with persistent state.