red

The Vim-style terminal editor where your coding agent works through your editor.

Vim muscle memory, editor-aware Codex agents, focused inline assistance, and modern code intelligence in one self-contained Rust binary. Red works without configuration; optional agent support requires an installed, authenticated Codex CLI.

Website · Download · Getting started · Documentation · Community

The current documented release is v0.7.0.

Install

Homebrew

brew install codersauce/tap/red

macOS and Linux

curl --proto '=https' --tlsv1.2 -fsSL https://getred.dev/install.sh | sh

The installer selects the correct macOS or x86_64 glibc Linux archive, verifies its published SHA-256 checksum, installs to ~/.local/bin, and runs Red's built-in self-check.

Windows

irm https://getred.dev/install.ps1 | iex

The PowerShell installer verifies the release checksum, installs to %LOCALAPPDATA%\Programs\Red\bin, and adds that directory to your user PATH.

To pin a release or choose another directory:

RED_VERSION=0.7.0 RED_INSTALL_DIR="$HOME/bin" \
  sh -c "$(curl --proto '=https' --tlsv1.2 -fsSL https://getred.dev/install.sh)"

You can also download a prebuilt archive or build from source. Red's editor, default configuration, themes, and plugins are bundled into the executable.

Agent support is optional. It requires Codex CLI 0.144.1 or newer and a completed codex login.

Why Red

• Keep your Vim muscle memory. Familiar modes, motions, operators, text objects, splits, and keyboard-first pickers work alongside Tree-sitter and LSP.
• Give your agent the real editor. Codex sees open buffers, unsaved changes, diagnostics, and selections. Its validated edits pass through Red's editor transactions and save paths instead of modifying files behind the editor.
• Ask right where you are. Inline assistance reviews, explains, or refactors an exact selection or enclosing function; code changes stay unsaved and undoable.
• Keep the whole workflow in your terminal. File search, language servers, a full Git workspace, plugins, themes, and sessions are bundled with Red.
• Reconnect without losing your work. On macOS and Linux, detachable sessions retain buffers, plugins, LSP state, and running agents across terminal or SSH disconnects.

New in v0.7.0

The v0.7.0 release adds:

• An agent that points to the code it means. Ask Agent to explain a subsystem, then follow links in its answer directly to source-anchored annotations. Choose the conversation's model and reasoning effort without changing your global Codex settings.
• Real Vim-style multi-cursor editing. Press Ctrl-n to select repeated occurrences or Ctrl-Up / Ctrl-Down for vertical cursors. Extend each selection with familiar motions and apply one Unicode-aware, undoable edit.
• Inline help that gets out of the way. Exact foreground edits can apply immediately while remaining unsaved; background and wider same-file edits always require explicit review. Keep the full history, inspect changes, or continue a discussion in Agent.
• Your files stay protected when another tool changes them. Clean buffers reload automatically. Dirty buffers keep both versions until you compare, reload, save elsewhere, or explicitly overwrite.
• A faster, more complete editing workspace. Browse large file trees, coordinate LSP and optional Copilot completions, and format supported pasted ranges.

First five minutes

Open a file:

red path/to/file

The first interactive launch offers a guided tour, release highlights, and an immediate exit into the editor. The tour uses a disposable practice buffer and safe Git/agent demonstrations; reopen it with :welcome or :tutorial. Starter configuration remains optional: embedded defaults, plugins, and themes are enough to begin editing.

See Getting started for editing, navigation, configuration, language servers, Git, CLI, and troubleshooting guidance. The Vim compatibility matrix is the precise, versioned behavior contract.

Agents that understand your editor

The full Agent workspace

1. Ask. Press Space A; Red provides a bounded source excerpt, current selection, relevant diagnostics, and authoritative unsaved buffer contents.
2. Work through the editor. Codex reads and changes files only through Red's workspace-confined, revision-checked editor tools. Agent edits are attributed to their conversation and saved to disk through Red.
3. Keep the conversation. Follow tool progress, queue another request, and resume the same conversation without losing editor context. Source-linked annotations turn explanations into navigable code walkthroughs.

In v0.7.0, file-tool playback is optional and disabled by default; set [agent] follow_tool_calls = true to reveal each target and pause before the operation. Full Agent writes still save to disk.

Focused inline assistance

Press Space i to review, explain, or refactor the enclosing function, falling back to the current line when syntax is unavailable. In Visual or Visual Line mode, the target is exactly the selected text. Requests use an ephemeral Codex thread with bounded read-only project context; visual-block targets are not supported. Inline code edits form one unsaved, undoable editor transaction.

An exact-target edit may apply immediately when its original popup is in the foreground; set [agent] auto_apply_inline_edits = false to review it first. Background results and wider same-file proposals always require an explicit review and approval. Use Space H to revisit inline history or A to prepare a full Agent follow-up without sending it automatically.

The agent workflow and safety contract explains prerequisites, path protections, exact boundaries, commands, and failure modes.

What Red ships today

The current release includes:

• Normal, Insert, Visual, Visual Line, Visual Block, and Command modes with expanding Vim motion and editing compatibility
• tree-sitter highlighting for Rust, Markdown, JavaScript, TypeScript/TSX, JSON, TOML, YAML, Bash, PowerShell, Lua, and Husk
• a first-party Husk language server plus built-in LSP defaults for Rust, TypeScript/JavaScript, Markdown, JSON, TOML, YAML, and Lua
• command and keymap discovery, fuzzy files, buffer navigation, symbols, references, project search, and diagnostics
• native Git gutter signs, hunk actions, and a full-screen workspace for staging, commits, Codex-generated commit messages, branches, remotes, stashes, logs, and rebases
• a persistent full Agent workspace and bounded inline assistance with source-anchored comments, retained history, and reviewable local edits
• an embedded Husk runtime with bundled file tree, search, Git, progress, inlay-hint, symbol, theme, and agent plugins
• a branded startup splash, the Red theme, accessible selection and cursor contrast, and optimized rendering hot paths
• atomic crash recovery on every platform and detachable sessions on macOS and Linux

See the latest release notes or the complete changelog for details.

After installing a new version, Red opens a themed What’s new panel once. Reopen it whenever you like with :whats-new, :changelog, or the command palette. Release notes are bundled for offline use and refreshed from the exact matching GitHub release in the background.

Configuration

Red layers your settings over embedded defaults, so a configuration file can contain only the values you want to change:

# ~/.config/red/config.toml
theme = "red.json"
scrolloff = 8
# Disable the automatic release panel or its optional GitHub refresh:
# show_whats_new = false
# fetch_release_notes = false

[search]
ignorecase = true
smartcase = true

[keys.normal]
"Ctrl-s" = "Save"

The commented default_config.toml documents every setting that ships with Red. Custom th