Eight Claude agents, one corner office, zero headcount
It exists to give small teams a persistent, multi-domain executive advisor by orchestrating eight specialist Claude agents behind one coherent voice that remembers past decisions.
SenteLabs built an eight-agent open-source council that remembers your decisions and nags you about follow-ups, offering a self-hosted counterweight to proprietary executive AI.

What it does OpenExecutive is a FastAPI and Next.js application that acts as a single senior executive advisor while routing questions to eight specialist Claude agents covering strategy, finance, HR, legal, operations, marketing, product, and board communications. The user never sees the internal seams; all responses are synthesized into one consistent voice. Beyond Q&A, it stores episodic memory of past decisions in SQLite and runs a built-in scheduler to proactively surface follow-ups.
The interesting bit
The architecture treats prompt caching as a first-class design constraint: the system prompt is split so the executive persona, company profile, and knowledge index stay cached separately, with the project reporting up to an 85 percent cache hit rate after the first few turns. Meanwhile, a background claude-haiku-4-5 pass extracts key decisions into SQLite after every response, so the next session opens with a <past_decisions> block—essentially giving the agent a corporate diary it actually reads.
Key highlights
- Eight specialist agents (CSO, CFO, CHRO, General Counsel, COO, CMO, CPO, Board Communications Director) are orchestrated in parallel by a
claude-sonnet-4-6executive persona. - RAG pulls from two ChromaDB collections: built-in MBA-level Markdown knowledge and your own uploaded company documents.
- Episodic memory persists across sessions via SQLite, with a background model summarizing key decisions after each turn.
- Multi-channel access: Web UI, Slack, Email, Telegram, Google Chat, Discord, and a Click CLI.
- The scheduler claims due actions with an
UPDATE … RETURNINGlock, which forces a single-instance deployment topology.
Caveats
- The scheduler relies on database-level row locking, so the API is explicitly designed to run as a single instance; horizontal scaling will double-fire scheduled actions unless you gate the scheduler first.
- The first local boot downloads roughly 90 MB of embedding models and heavy ML dependencies, so initial startup is slow; subsequent starts are fast.
Verdict Founders and small-team operators who want a structured sounding board for strategy, finance, and ops will get the most value; if you already have a full C-suite, you are not the target audience.
Frequently asked
- What is SenteLabsAI/OpenExecutive?
- It exists to give small teams a persistent, multi-domain executive advisor by orchestrating eight specialist Claude agents behind one coherent voice that remembers past decisions.
- Is OpenExecutive open source?
- Yes — SenteLabsAI/OpenExecutive is an open-source project tracked on heatdrop.
- What language is OpenExecutive written in?
- SenteLabsAI/OpenExecutive is primarily written in Python.
- How popular is OpenExecutive?
- SenteLabsAI/OpenExecutive has 4.1k stars on GitHub and is currently cooling off.
- Where can I find OpenExecutive?
- SenteLabsAI/OpenExecutive is on GitHub at https://github.com/SenteLabsAI/OpenExecutive.