A terminal chat client that actually talks to your editor

What it does

Oatmeal is a Rust-built terminal UI for chatting with large language models through pluggable backends. It wraps OpenAI, Ollama, Claude, Gemini, and LangChain into a single interface with slash commands, chat bubbles, and persistent sessions. The standout trick is its editor integration: it can read selected code from Neovim, send it to a model, and push the resulting code blocks back into your buffer.

The interesting bit

Most terminal LLM clients force you to copy-paste code across windows; Oatmeal wires the chat directly into your editor. Model responses tag each code block with an index, letting you append or replace specific snippets in Neovim without leaving the terminal. That small plumbing detail turns chat from a side task into a coding flow state.

Key highlights

• Backend-agnostic: swap between cloud APIs and a local Ollama instance on the fly.
• Editor-aware workflow: supports Neovim via a companion plugin, plus clipboard and “none” modes.
• Code actions: /append, /replace, and /copy target individual or ranges of indexed code blocks.
• Session persistence: old conversations are stored and can be reopened or deleted later.
• Syntax highlighting via embedded themes or any Sublime Text/TextMate .tmTheme file.

Caveats

• The README warns the project is young, LLMs can return unexpected output the UI isn’t ready for, and bugs are likely hiding.
• Claude, Gemini, and LangChain backends are marked experimental.
• Searching across past sessions requires external tools like Ripgrep; native session search isn’t implemented yet.

Verdict

Neovim users and terminal-native developers who want LLM assistance wired directly into their editing flow should take a look. If you prefer a polished GUI or don’t need editor round-tripping, you can safely skip it.