Zero-Trust AI Agent
零信任 AI 代理運行時
High-performance. Single binary. Sandboxed by default. Multi-mode: CLI agent, CI resource, and AI-native Data Exchange Hub (Rune Notes).
高性能 · 單一獨立執行檔 · 預設嚴格沙盒隔離 · 多模式運作:命令列 Agent、CI 流程資源與 AI 原生資料交換中心(Rune Notes)。
cgroups v2, Network namespace (or net-guard), Seccomp BPF, Landlock, DNS allowlist. Every tool call is isolated.
cgroups v2、網路命名空間(或 net-guard)、Seccomp BPF、Landlock、DNS 白名單。隔離每一次工具調用。
read_file, write_file, list_dir, execute_cmd, fetch_url — all sandboxed.
read_file、write_file、list_dir、execute_cmd、fetch_url — 全部受沙盒嚴格管控。
Load contextual abilities via @skill_name. Frontmatter-driven, multi-path search.
透過 @skill_name 動態載入專屬技能。支援 Frontmatter 配置與多路徑搜尋。
GitHub Copilot (auto refresh), OpenRouter, Google Gemini, any OpenAI-compatible endpoint.
支援 GitHub Copilot(自動刷新憑證)、OpenRouter、Google Gemini 與任何相容 OpenAI 的 API 端點。
Stdio JSON-RPC client for Model Context Protocol servers. Extend without recompiling.
內建 Stdio JSON-RPC Model Context Protocol 客戶端,無需重新編譯即可無縫擴展第三方工具。
Same binary doubles as a Concourse CI resource type via symlink routing. Zero extra deps.
同一二進位檔可透過符號連結直接作為 Concourse CI 資源類型運作,零額外依賴。
Adjustable reasoning depth: off, low, medium, high, xhigh. Compatible with Claude and o-series models.
可自由調整推理深度:off、low、medium、high、xhigh。相容 Claude 與 o 系列推理模型。
Embedding-powered skill discovery. Skills matched by meaning, not just @name references.
以向量嵌入(Embedding)驅動技能發現,依據語意自動匹配合適技能,不限於純名稱引用。
Strace-based exploratory analysis that suggests safe policy expansions; human-reviewed before applying.
基於 strace 探索分析工具行為並主動建議安全的權限擴展,經人工審查確認後生效。
Override the default system prompt via CLI flag, env var, or config. AGENTS.md context is always appended.
可透過 CLI 參數、環境變數或設定檔自訂系統提示詞,並始終自動注入 AGENTS.md 安全上下文。
AI-native Markdown Data Exchange Hub served from the same single binary. Exposes Web UI, MCP endpoint, and REST API.
由同一二進位檔直接啟動的 AI 原生 Markdown 資料交換中心,提供 Web UI、MCP 端點與 REST API。
First-party browser extension for Chrome & Firefox side panel. Chat about active pages and update shared notes directly.
Chrome 與 Firefox 側邊欄專屬擴充套件,能針對當前瀏覽網頁進行 AI 對話並直接記錄至筆記中樞。
Native FIM ghost text completion and interactive commands (:RuneAsk, :RuneEdit, :RuneStatus, :RuneLog) built in via --features vim.
以 --features vim 內建原生 FIM 幽靈文字補全與互動式指令(:RuneAsk、:RuneEdit 等)。
A responsive web app and central Markdown data exchange hub served from a single self-contained binary. One process serves the web UI, the MCP endpoint, and the REST API.
由單一獨立執行檔直接啟動的響應式 Web 應用程式與 Markdown 資料交換中樞。單一處理程序即可同時提供 Web UI、MCP 端點與 REST API。
Zero runtime dependencies on Linux. A single process serves the full web UI, built-in AI assistant, MCP server endpoint (POST /mcp), and REST API.
在 Linux 下完全零依賴。單一程序即可提供完整 Web UI、內建 AI 助理、MCP 伺服器端點(POST /mcp)以及 REST API。
External agents (Antigravity, OpenClaw, Hermes, Copilot, OpenCode) connect via Model Context Protocol to read, search, and edit notebooks in real-time.
外部 Agent(Antigravity、OpenClaw、Hermes、Copilot、OpenCode)可透過 Model Context Protocol 即時讀取、搜尋與編輯筆記本。
First-party extension for Chrome & Firefox side panel. Discuss the current browsing page and push notes directly into the hub over REST API.
Chrome 與 Firefox 側邊欄官方擴充套件,能針對當前網頁即時討論並透過 REST API 將重點寫入筆記中樞。
Runs seamlessly across Desktop (3-column layout), Tablet (collapsible panels), and Mobile (adaptive single-column) without separate installations.
無縫支援桌面端(三欄版面)、平板端(摺疊面板)與行動手機(自適應單欄),免額外安裝應用程式。
KaTeX math ($E=mc^2$), Mermaid diagrams-as-code, raw inline SVG markup, and syntax highlighting. Agents generate diagrams purely as text.
支援 KaTeX 數學式($E=mc^2$)、Mermaid 架構圖即程式碼、原生行內 SVG 與語法高亮。Agent 可純以文字生成清晰圖表。
GitHub OAuth 2.0, standard OIDC (Google/Okta/Authentik), and standalone local accounts. Granular public/private sharing for individual files.
支援 GitHub OAuth 2.0、標準 OIDC(Google / Okta / Authentik)與本機獨立帳號。個別檔案可自訂公開或私有存取權限。
# Run locally
rune notes --bind 0.0.0.0 --port 9527
# Or run with Docker
docker run -d --name rune-notes \
-v ~/.rune:/home/rune/.rune \
-v "$PWD"/notes:/notes \
-p 9527:9527 \
ghcr.io/fourdollars/rune notes --bind 0.0.0.0 --port 9527
Use Rune as a Concourse CI resource type. Minimal weather pipeline:
將 Rune 作為 Concourse CI 資源類型使用。簡易天氣檢查 Pipeline 範例:
resource_types:
- name: rune-agent
type: registry-image
source:
repository: ghcr.io/fourdollars/rune
tag: latest
resources:
- name: weather
type: rune-agent
check_every: 1h
source:
api_key: ((copilot-pat))
model: gpt-4o-mini
prompt: "Fetch the weather for Taoyuan from wttr.in using curl."
policy:
allowed_commands: ["curl"]
allowed_domains: ["wttr.in"]
jobs:
- name: weather-check
plan:
- get: weather
trigger: true
Run Rune instantly via Docker. Three image variants available:
透過 Docker 立即執行 Rune,提供三種映像檔版本:
# Alpine — smallest image (~15 MB), musl-based
docker pull ghcr.io/fourdollars/rune:alpine
# Debian (Bookworm) — glibc-based, broad compatibility
docker pull ghcr.io/fourdollars/rune:debian
# Ubuntu (24.04) — familiar environment
docker pull ghcr.io/fourdollars/rune:ubuntu
# Latest tag (defaults to Debian)
docker pull ghcr.io/fourdollars/rune:latest
# Initialize config — mounts ~/.config/rune for persistence
docker run -it --rm \
-v ~/.config/rune:/home/rune/.config/rune \
ghcr.io/fourdollars/rune init
# Start an interactive session
docker run -it --rm \
-v ~/.config/rune:/home/rune/.config/rune \
-v "$PWD":/workspace -w /workspace \
ghcr.io/fourdollars/rune
# One-shot query without local installation
echo "Explain what Landlock is" | docker run -i --rm \
-v ~/.config/rune:/home/rune/.config/rune \
ghcr.io/fourdollars/rune --json --yes
# Add to ~/.bashrc or ~/.zshrc
alias rune='docker run -it --rm \
-v ~/.config/rune:/home/rune/.config/rune \
-v "$PWD":/workspace -w /workspace \
ghcr.io/fourdollars/rune'
git clone https://github.com/fourdollars/rune.git
cd rune
cargo build --release
cp target/release/rune ~/.local/bin/
rune init
echo "Get weather for Tokyo" | rune --json --yes
Zero-dependency native Vim 9+ & Neovim integration compiled into the binary.
零依賴、編譯進執行檔的原生 Vim 9+ 與 Neovim 整合插件。
cargo build --release --features vim
rune vim install
:RuneAsk <prompt> " Ask AI questions (opens side markdown view)
:RuneExplain " Detailed code explanation
:RuneEdit <prompt> " Refactor or edit current buffer
:RuneFix " Auto fix bugs & issues
:RuneStatus " Inspect provider, model & rate budget
:RuneLog " Open side debug window (__Rune_Log__)
:RuneToggle(!) " Toggle inline completion on/off
:RuneEnable(!) " Enable inline completion
:RuneDisable(!) " Disable inline completion
set statusline=%f\ %m%r%=%{rune#statusline()}\ [%{rune#model()}]