What is opsintech-platform?

OpsinTech Platform is an AI agent platform built for companies that want to run AI agents in production, not just as demos. It is built on top of DeerFlow, an existing open source agent runtime from ByteDance that already has a large following, and adds the management and safety features that a company needs before letting many people use agents at once. The main addition is multi tenancy: instead of one person running the agent tools alone, OpsinTech separates data and settings into three layers, user, tenant, and global, so different teams or customers stay isolated from each other. On top of that sits role based access control, with three roles, platform admin, tenant admin, and tenant member, each with different permissions. When someone registers, a personal tenant is created for them automatically, so there is no manual setup step. Configuring which AI models the agents can use is handled through an admin dashboard rather than editing configuration files by hand. The platform ships with more than 20 ready made provider templates, covering services such as OpenAI, Anthropic, DeepSeek, Gemini, Ollama, and several others, so a user can pick a provider from a dropdown instead of writing code. Models are stored in a database and can be turned on, marked deprecated, or retired per tenant. For governance, every operation performed on the platform is recorded in an audit trail that stays separated by tenant, users can be suspended, and new accounts are required to change their password on first login. The admin interface is available in English, Chinese, Japanese, and Korean, and detects the browser's language automatically. To run OpsinTech, a user needs Docker, since the whole platform starts with a single docker compose command. The backend is written in Python 3.12 or newer and the project also uses Node.js 22 or newer, with PostgreSQL as the supported database. The README says the project's longer term goal is to grow beyond agent chat into handling real operations tasks like alerts, incidents, and workflows, built around a sequence the maintainers describe as discover, understand, decide, execute, and audit. It is released under the MIT license.