Vermilion Editor
A small IDE for small people.
A native Linux code editor with a built-in AI agent, build and debug support, and a live model monitor (CTXMon) for local llama.cpp servers. One window for the whole loop: edit, ask the agent, review what it changed, build, run, debug, and watch the model serving it.
| Language | C++17 |
|---|---|
| Toolkit | GTK 4, GtkSourceView 5 (plain C APIs) |
| Build | CMake |
| Platform | Linux (Ubuntu 24.04 reference), Wayland & X11 |
| Licence | GPL-3.0-or-later |
| Status | Pre-alpha. Milestone M0 (foundation) is done; M1 (CTXMon monitor logic) is under way. |
Details
- No language bindings or vendor SDKs — the agent speaks provider HTTP protocols directly; builds and debuggers run as subprocesses.
- Planned agent support for OpenAI-compatible servers, Anthropic Messages API and Gemini, with Restricted / Balanced / Unrestricted / Custom permission presets.
- Every agent turn is a reviewable change set with side-by-side diff and per-hunk accept or revert.
- CTXMon: a native C++ port of a Python llama.cpp monitor — context usage, token speeds, graphs and VRAM, polled from /slots and /props.
- Debugger via the Debug Adapter Protocol (gdb first) and an MCP server/client planned for later releases.
- Logging with secret masking and a growing unit test suite verified against the Python original.