# Open Executive

> An open-source AI system that acts as a virtual executive team, providing a single coherent executive voice backed by eight specialist Claude agents covering strategy, finance, HR, legal, operations, marketing, product, and board communications.

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.

## Features
- Eight specialist AI agents (CSO, CFO, CHRO, GC, COO, CMO, CPO, Board)
- Single coherent executive voice synthesized from parallel specialist calls
- RAG with two ChromaDB layers: built-in MBA knowledge and company documents
- Episodic memory via SQLite across sessions
- Built-in proactive scheduler for follow-ups and time-sensitive actions
- Anthropic prompt caching (up to 85% cache hit rate)
- Web UI, Slack, Email, Telegram, Google Chat, Discord, and CLI interfaces
- Document upload (PDF, XLSX, Markdown) via UI, API, or CLI
- Local model support (Ollama, LM Studio, vLLM, llama.cpp)
- OpenRouter integration for non-Anthropic models
- Company onboarding wizard for profile setup
- LLM-as-judge evaluation system with 29 scenarios
- Google sign-in with email allow-list access control
- Fly.io deployment with dev and QA environments
- Apache 2.0 license — free for commercial use

## Integrations
Anthropic Claude API, Slack, Discord, Telegram, Google Chat, Gmail (IMAP/SMTP), ChromaDB, SQLite, Ollama, LM Studio, vLLM, llama.cpp, OpenRouter, Honcho (optional per-person memory), Fly.io, Google OAuth

## Platforms
WEB, API, CLI

## Pricing
Open Source

## Version
main

## Links
- Website: https://github.com/SenteLabsAI/OpenExecutive
- Documentation: https://github.com/SenteLabsAI/OpenExecutive/blob/main/docs/architecture.md
- Repository: https://github.com/SenteLabsAI/OpenExecutive
- EveryDev.ai: https://www.everydev.ai/tools/open-executive