gptme

gptme is an AI agent for developers and people doing data analysis or other knowledge work in their own environment. It can act on files and use installed tools, with a choice of local models or cloud providers.

• Coding and analysis: Edit files, execute shell commands and Python, run tests and work with GitHub, with session state retained between steps.
• Browser and desktop tasks: Search the web, interact with pages and read PDFs through Chromium, analyze images and screenshots, and control desktop apps on macOS or X11.
• Persistent agents: Keep a Git-tracked journal, memory and task queue, run on schedules or events, and assign separate work to child agents.
• Editor and tool connections: Use gptme in Zed or JetBrains through ACP, load external MCP tools, or expose its shell, Python and file tools to clients such as Cursor and Claude Desktop.
• Reusable workflows: Add tools through plugins and capture instructions in Markdown skills and lessons, with a local memory store that can also serve Claude Code and Codex.

The software is free and open source under MIT, with no gptme account required for self-hosted use. It runs on laptops, remote servers and in CI; a desktop app supports Linux, macOS and Windows. Its web UI can connect to your own server.

Local backends include Ollama, LM Studio, vLLM and llama.cpp, plus OpenAI-compatible servers. Local inference runs on your hardware; cloud models such as Claude, GPT, Gemini, Grok and DeepSeek receive requests at the chosen provider. Browser tasks and external tools can still contact network services.