Rubra

What is Rubra?

Rubra is an open-source platform for creating AI assistants that run entirely on your own machine. Because all processing happens locally, you avoid costly API tokens, keep chat histories and data private, and still get an efficient chat interface for working with models. Rubra can also handle data from different sources.

Compared to OpenAI’s ChatGPT, Rubra is built for local development: it prioritizes privacy, ships with built-in LLMs, and lets you switch between local and cloud setups.

It includes several pre-trained models optimized for local use — Mistral-7B, Phi-3-mini-128k, and Qwen2-7B — improvements on well-known open-source LLMs. You can also connect to models from OpenAI and Anthropic, choosing whichever model suits your task.

The platform offers an API compatible with OpenAI’s Assistants API, so work can move between your local machine and the cloud. Rubra is an open project: you can contribute code, report bugs, and give feedback through its GitHub repository. Developers can test AI assistants locally and see how they perform in real-world settings while keeping data private.

Who created Rubra?

The Rubra team launched the platform as an open-source tool for building AI assistants that run locally. The goal: an affordable, private way for developers to create AI-powered applications using large language models (LLMs) on their own computers, without relying on API tokens.

What is Rubra used for?

• Building AI assistants that run on your local machine
• Creating AI-powered applications affordably and privately, without API tokens
• Comparing how different AI models perform
• Calling external tools you define in a predictable, deterministic way while the AI reasons and chats — useful for agentic tasks where the AI needs to act independently
• Improving open-weight LLMs by adding tool-calling capabilities, which can also help prevent models from “forgetting” what they’ve learned
• Switching between local and cloud development, or contributing code and ideas to the project

Key takeaway

Rubra suits developers who want to build and test AI assistants locally, keep costs and data exposure low, and retain the flexibility to move to cloud models when needed.