A desktop reviewer crafted for command-line invocation. Escape the flood of terminal text with a single shortcut: read long AI replies comfortably in rich text, make precise selection-based annotations, and write the aggregated prompt right back to your target workflow.
Full animated preview for GitHub. Click to open the MP4 demo on GitHub.
AI coding agents (Codex, Claude Code, Gemini CLI) produce long, structured replies — but you're reading them in a terminal. That's fine for 20 lines, painful for 500.
cliV can be set as an external editor for Claude Code, Codex, and Gemini, invoked via shortcut, enabling review, annotation, prompt return, and history functions:
- Review — full Markdown + Mermaid diagram rendering, no more plain text
- Annotate — select exact passages to add comments instead of writing vague follow-ups
- Write back — write back to the active write target when available, fall back to clipboard otherwise
- Open — also open local Markdown files for standalone review
Hook commands, path rules, launch behavior, and debugging are documented in the guide links below.
Please use the dedicated guide documents:
What each guide covers:
Install Guide: official install paths,$EDITOR, basic hook setup, verification, and uninstallAdvanced Integration & Debugging Guide: non-default executable paths, launch semantics, reply-cache lookup rules, and manual debugging
For source builds and developer workflows, see the additional documents under docs/ and the contributor notes in AGENTS.md.
- Start your AI agent (Codex, Claude Code, or Gemini CLI)
- Have a conversation that generates a long reply
- Trigger the agent's
$EDITORflow (commonlyCtrl+G, but depends on agent / config) - cliV opens with the agent's latest reply rendered as rich Markdown
- Select text → annotate → aggregate → write back or copy the result
%%{init: {"themeVariables": {"fontSize": "18px"}, "flowchart": {"nodeSpacing": 60, "rankSpacing": 80, "padding": 24}} }%%
flowchart LR
subgraph A["CLI Coding Tools"]
A1["Claude Code"]
A2["Codex"]
A3["Gemini"]
end
B["Trigger `$EDITOR` via shortcut<br/>(e.g., Ctrl+G)"]
subgraph C["cliV Review Phase"]
C1["Auto-extract latest reply"]
C2["Rich text reading"]
C3["Add/Edit/Delete annotations"]
C4["Aggregate annotations as Prompt"]
end
subgraph D["Return Result"]
D1["Write back to explicit target"]
D2["Fallback to clipboard"]
end
A1 --> B
A2 --> B
A3 --> B
B --> C1 --> C2 --> C3 --> C4
C4 --> D1
C4 -. No write target .-> D2
classDef cli fill:#EDF4FF,stroke:#5B8DEF,color:#17325C,stroke-width:1.5px;
classDef review fill:#EEF8F1,stroke:#43A047,color:#1F4D2E,stroke-width:1.5px;
classDef back fill:#FFF4E5,stroke:#FB8C00,color:#6A3A00,stroke-width:1.5px;
class A1,A2,A3,B cli;
class C1,C2,C3,C4 review;
class D1,D2 back;
- 📖 Rich Markdown rendering — headings, code blocks, tables, Mermaid diagrams
- ✏️ Selection-based annotations — highlight passages and add comments (in-text highlights rely on CSS Highlight API)
- 📋 Write-back flow — aggregate annotations into a prompt, then write back or copy
- 🔄 Multi-agent support — best-effort auto-detection of Codex / Claude / Gemini; force with
CLIV_AGENT - 📂 Open local Markdown — review cached replies or open
.mdfiles directly with safe review-only defaults - 🎛️ Unified settings — manage Reading, Prompts, and Shortcuts from one settings surface backed by
~/.cliv/config.toml
- Frontend: React 19 + Vite 7 + Zustand + TailwindCSS 4
- Backend: Tauri v2 (Rust)
- Rendering: react-markdown + remark-gfm + Mermaid
- Virtual scrolling for large documents
- Diff / suggestion mode
- Standalone review polish (
cliv <file.md>now stays review-only by default) - Cross-platform builds (macOS, Windows)
- Plugin system for custom agents
- Review history (project-grouped archives with read-only replay)
- Favorites
- Iterative editing mode
- Linux DO — Learn AI, go LinuxDO!
MIT