Repository navigation
Home
AI-powered system administration. One binary. No installation required.
Kelpy is an AI assistant for system administrators, shipped as a single executable. Copy it onto a server, run it,
do the work with it, then stop it: nothing of Kelpy keeps running. To download and start it, run
curl -fsSL https://kelpy.io/install.sh | sh on Linux or irm https://kelpy.io/install.ps1 | iex in PowerShell.
The runtime, the local server and the Web UI are in one file, with no Node.js, Python, Docker or external database.
Kelpy is in early development. It runs the Web UI with authenticated access and chats with the model you choose, with streamed answers and conversations that survive a restart. With your approval it looks at the system and changes it: files, services, packages and databases through operations it shows first and checks afterwards, anything else with commands you approve. It keeps a memory of the server and tells you what was changed outside it. The terminal UI is planned.
- Installation and first run: the install command, the Kelpy home, the first run, reaching the Web UI locally or through SSH.
- Connecting to AI: a ChatGPT plan, a free online service, an API key or a local model.
- System access: what Kelpy may look at and change, when it asks, the Command log and the memory.
- Command line: commands and options.
- Security and data: what Kelpy stores, where, and who can reach it.
- Troubleshooting: messages you may see, and what to do.
Kelpy keeps what it knows about a server in the memory/ folder of its home, as Markdown pages you can read and
edit. The model receives memory/MEMORY.md with every answer, reads the other pages when it needs them and notes
lasting facts as it works. server.md is written by Kelpy from its records: what it changed, and what was changed
outside it. See System access.
Kelpy's memory management is strongly inspired by the MEM project.
Kelpy is licensed under either the Apache License 2.0 or the MIT license, at your option.