OpenGhost
Open desktop AI agent for Windows, macOS and Linux, written from scratch: its own agent engine, its own browser tooling and its own drawing engine that shows wh
What is OpenGhost?
Open desktop AI agent for Windows, macOS and Linux, written from scratch: its own agent engine, its own browser tooling and its own drawing engine that shows what it explains.
Overview of OpenGhost
What it is
OpenGhost is an open desktop AI agent for Windows, macOS and Linux, written from scratch. It includes its own agent engine, browser tooling, and drawing engine. The agent runs commands, edits files, keeps git, and works on the web in a built-in browser. It visualizes explanations with charts, schemes, photos, and videos. The app is built on Electron and Anthropic's official SDK.
Key features
- Agent engine with tools for commands, files, git, web search, pages, PDFs, and video frames, plus three permission modes.
- Drawing engine that renders 43 kinds of drawings, including flowcharts, loss curves, meal plans, and trip routes; drawings are interactive and editable.
- Built-in browser with its own cursor, capable of clicking, typing, and reading pages; works within user accounts via a custom MCP server.
- Markdown renderer, code highlighter, and math support, all custom-built.
- Notes feature: the agent reads and remembers notes, and proactively brings them up when relevant, but never acts without permission.
- Memory system: remembers user preferences across chats, with editable records and an on/off switch.
- Quick chat (Alt+Space) and mini chat for parallel tasks, with agents that continue working when closed.
- Chat stats showing token usage, model share, cache usage, and context fullness; compact chat frees context.
- Supports attaching photos, videos, and PDFs; videos are analyzed frame by frame, PDFs read as text and images.
Use cases
- Learning and exploration: ask how something works and receive visual explanations with drawings.
- Project work: manage files, run commands, and use git within a project folder.
- Web automation: interact with websites, including personal accounts, for tasks like checking post views or reading comments.
- Document analysis: read PDFs, extract information, and summarize content.
- Personal assistance: set reminders via notes, get answers in quick chats from anywhere.
Getting started
See the repository README for installation instructions. The app is available for Windows, Linux, and macOS, with a browser-based trial on the website.
License and pricing
The license is MIT, and the project is open source. Pricing is not stated; it is open source, so no pricing applies.
Based on the project's public README and repository metadata.