OpenCompany

OpenCompany is a self-hosted system for running AI employees: long-running agent teams that each own a business job, connect to your email, calendar, messages, code, and other tools, and work in the background on your own machine. It ships as a desktop app for Windows, macOS, and Linux, a headless server install, or a Docker image.

An AI employee is a small team rather than a single chatbot. A lead agent understands the job, hands work to specialist agents, checks their results, and reports back to you. You hire an employee by describing the job in plain language, and OpenCompany drafts its setup (the apps it uses, when it works, and what it checks with you first). In Dev mode you build or adjust the team on a visual canvas by dragging in agents and connecting tools, without code. Employees wake on events such as a new email, a customer message, or a schedule, and can be set to show each reply for approval before it goes out.

Ready-made employees include a Builder (code, pull requests, and deployments), a Grower (publishing, community, and market analysis), a Chief of Staff (mail, calendar, and daily summaries), a Front Desk (customer replies across chat channels), a Researcher (browsing and a shared knowledge base), and a Treasurer (Stripe payment operations).

Employees improve without retraining a model. They keep an editable memory of facts and decisions, continue the same conversation across tasks and summarize it when it grows long, add tools or skills from your library within the rules you set, and follow skills written as plain-text playbooks that you can edit. Everything an employee learns stays readable in its memory, notes, skills, and canvas.

OpenCompany works with models from OpenAI, Anthropic, Google, xAI, DeepSeek, Kimi, Mistral, Groq, Cerebras, Sarvam, and OpenRouter, or local models through Ollama, LM Studio, or any OpenAI-compatible server such as llama.cpp or vLLM. Its site compares it with workflow tools such as n8n and with agent runtimes such as OpenClaw, Hermes Agent, and Paperclip, and describes it as an organization layer built on top of an agent runtime.