gitmyhub

polyrob

Python ★ 3 updated 9h ago

A self-hosted autonomous AI agent that pursues goals, learns from experience, and runs on your own machine.

A self hosted AI agent that runs on your own machine, plans and carries out goals, and learns from past sessions.

PythonSQLitePlaywrightPyPIsetup: moderatecomplexity 4/5

Polyrob is a self hosted AI agent that runs on your own computer instead of a company's cloud service. You give it a goal in plain language, and it plans the steps needed, browses the web, reads and writes files, runs code and shell commands, calls outside tools and APIs, and tries to recover on its own if something goes wrong along the way.

The project supports every major AI provider, including OpenAI, Anthropic, Google, DeepSeek, OpenRouter, and NVIDIA NIM, and can automatically switch to a backup provider if one fails or is rate limited. It remembers what it learns between sessions and can be reached through several places: a terminal, a web dashboard, a REST API, or chat apps like Telegram, email, WhatsApp, Discord, Slack, Signal, and X.

A personal agent mode turns Polyrob into something more durable. In this mode it keeps a backlog of goals that survives a restart, schedules its own recurring tasks using plain language like every monday at 9am, wakes itself up to check on unfinished work, and writes new skills based on what worked before. Skills follow an open format also used by Claude Code, so they can be installed from a folder, a GitHub repo, or a direct link, with a review step before anything new takes effect.

For tools, it can fetch web pages as clean text, or use full browser automation for clicking and navigating real sites. It can connect to Model Context Protocol servers, pull structured data from many websites, handle files and code, and even reason over screenshots and images.

The project requires Python 3.11 or later and installs with pipx. It is released under the MIT license, which allows free use including commercial use. The full README is longer than what was shown.

Where it fits