Open Executive

Open Executive is an open-source project by SenteLabsAI that gives companies a virtual executive team powered by Anthropic's Claude models. It presents a single, consistent executive persona to the user while internally routing queries to eight specialist AI agents — each covering a distinct C-suite domain. The project is licensed under Apache 2.0 and self-hostable, with a live demo environment deployed on Fly.io.

What It Is

Open Executive is a multi-agent AI system built with Python (FastAPI) on the backend and Next.js 15 on the frontend. Rather than exposing the internal agent architecture, it synthesizes responses from up to eight parallel specialist calls into one coherent executive voice. The eight specialists are:

• Chief Strategy Officer — competitive analysis, M&A, market positioning, OKRs
• Chief Financial Officer — financial modeling, fundraising, unit economics, cash flow
• Chief HR/People Officer — hiring, compensation, performance, culture
• General Counsel — contracts, IP, employment law basics, compliance Chief Operating Officer — process design, vendor management, operational scaling
• Chief Marketing Officer — GTM strategy, brand, communications, PR
• Chief Product Officer — roadmap, prioritization, product strategy
• Board Communications Director — board decks, investor relations, governance

Architecture and Knowledge Layers

The orchestrator runs on claude-sonnet-4-6 and dispatches specialist calls in parallel via tool use. Deep-reasoning specialists (CSO, CFO, GC, Board) use claude-opus-4-7 with extended thinking. Each specialist retrieves context from two ChromaDB layers: built-in MBA-level Markdown knowledge seeded at startup, and company-specific documents uploaded by the operator. RAG context is injected into the user turn rather than the cached system prompt, preserving Anthropic's prompt caching — the README notes up to an 85% cache hit rate after the first few turns.

Episodic memory is handled by a background claude-haiku-4-5 pass that extracts key decisions and initiatives into SQLite after every response. The next session opens with a <past_decisions> block so the Executive retains context across conversations. A built-in scheduler surfaces proactive follow-ups and time-sensitive actions, using UPDATE … RETURNING to prevent double-firing; the API is designed to run as a single instance.

Interfaces and Integrations

Open Executive ships with multiple access channels out of the box:

• Web UI at localhost:3000 (Next.js 15 + Tailwind)
• Slack — mention @OpenExecutive or DM the app
• Email — IMAP/SMTP poller for CC or direct email
• Telegram — configured bot token
• Google Chat — space mention
• Discord — DM, @mention, or /ask and /today slash commands
• CLI — openexecutive chat

Document upload (PDF, XLSX, Markdown, etc.) is supported via the web UI, API, or CLI, and uploaded files are chunked into a separate company_docs ChromaDB collection.

Local Model and OpenRouter Support

Beyond the default Anthropic API backend, Open Executive supports running against any OpenAI-compatible local server — Ollama, LM Studio, vLLM, or llama.cpp — by setting LOCAL_MODELS_ENABLED=true and pointing LOCAL_BASE_URL at the local server. Model slugs appear in a Council UI dropdown, enabling hybrid setups where the Executive persona stays on Claude while individual specialists run locally. OpenRouter routing is also supported via OPENROUTER_ENABLED, unlocking non-Anthropic models per agent. The README notes that prompt caching and extended thinking are automatically disabled for local models, and recommends tool-use-capable models like Llama 3.3 70B or Qwen2.5 for reliable multi-agent routing.

Evaluation System and Contribution Model

The repository includes 29 eval scenarios covering all eight domains, scored by claude-opus-4-7 as an LLM-as-judge. Each scenario defines a query, simulated company context, expected topics, required specialist routing, and a domain-specific rubric across five dimensions (persona coherence, domain accuracy, company context utilization, routing quality, actionability), each rated 1–5. The CI gate requires a ≥ 3.5/5 average; any dimension dropping more than 10% versus main fails the pull request. Contributors adding new specialist agents must include at least two eval scenarios, a system prompt, and registration in the router and knowledge retriever.

Current Status

The repository was created in June 2026 and last pushed in August 2026, with 1,577 GitHub stars and 107 forks as reported by the GitHub metadata. The project is actively maintained under Apache 2.0, with CI/CD pipelines deploying to separate dev and QA environments on Fly.io. A live demo UI is accessible at openexec-ui-dev.fly.dev behind Google sign-in with an email allow-list.