{"id":45197270,"url":"https://github.com/ghostapp-ai/ghost","last_synced_at":"2026-02-27T21:18:32.530Z","repository":{"id":339605579,"uuid":"1162644577","full_name":"ghostapp-ai/ghost","owner":"ghostapp-ai","description":"👻 Private local-first AI assistant for your desktop — hybrid semantic + keyword search, native AI inference, zero telemetry. Open-source Raycast/Spotlight alternative for Windows, macOS \u0026 Linux.","archived":false,"fork":false,"pushed_at":"2026-02-20T17:44:04.000Z","size":1421,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-02-20T18:46:41.276Z","etag":null,"topics":["ai","cross-platform","desktop-app","embeddings","llm","local-first","mcp","ollama","open-source","privacy-first","productivity","raycast-alternative","react","rust","semantic-search","spotlight-alternative","sqlite","tauri","typescript","vector-search"],"latest_commit_sha":null,"homepage":null,"language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ghostapp-ai.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":".github/FUNDING.yml","license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":".github/CODEOWNERS","security":"SECURITY.md","support":".github/SUPPORT.md","governance":null,"roadmap":"ROADMAP.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null},"funding":{"github":["AngelAlexQC"]}},"created_at":"2026-02-20T14:18:33.000Z","updated_at":"2026-02-20T17:44:07.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/ghostapp-ai/ghost","commit_stats":null,"previous_names":["ghostapp-ai/ghost"],"tags_count":30,"template":false,"template_full_name":null,"purl":"pkg:github/ghostapp-ai/ghost","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ghostapp-ai%2Fghost","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ghostapp-ai%2Fghost/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ghostapp-ai%2Fghost/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ghostapp-ai%2Fghost/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ghostapp-ai","download_url":"https://codeload.github.com/ghostapp-ai/ghost/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ghostapp-ai%2Fghost/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29914438,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-27T19:37:42.220Z","status":"ssl_error","status_checked_at":"2026-02-27T19:37:41.463Z","response_time":57,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","cross-platform","desktop-app","embeddings","llm","local-first","mcp","ollama","open-source","privacy-first","productivity","raycast-alternative","react","rust","semantic-search","spotlight-alternative","sqlite","tauri","typescript","vector-search"],"created_at":"2026-02-20T13:01:27.496Z","updated_at":"2026-02-27T21:18:32.525Z","avatar_url":"https://github.com/ghostapp-ai.png","language":"Rust","funding_links":["https://github.com/sponsors/AngelAlexQC"],"categories":[],"sub_categories":[],"readme":"\u003cp align=\"center\"\u003e\n  \u003cimg src=\"public/ghost-logo.svg\" width=\"120\" alt=\"Ghost Logo\" /\u003e\n\u003c/p\u003e\n\n\u003ch1 align=\"center\"\u003eGHOST\u003c/h1\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003eThe Private Agent OS for Desktop \u0026 Mobile\u003c/strong\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"#features\"\u003eFeatures\u003c/a\u003e \u0026bull;\n  \u003ca href=\"#architecture\"\u003eArchitecture\u003c/a\u003e \u0026bull;\n  \u003ca href=\"#getting-started\"\u003eGetting Started\u003c/a\u003e \u0026bull;\n  \u003ca href=\"#roadmap\"\u003eRoadmap\u003c/a\u003e \u0026bull;\n  \u003ca href=\"#contributing\"\u003eContributing\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://github.com/ghostapp-ai/ghost/releases/latest\"\u003e\u003cimg src=\"https://img.shields.io/github/v/release/ghostapp-ai/ghost?style=flat-square\u0026color=blue\" alt=\"Release\" /\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/ghostapp-ai/ghost/blob/main/LICENSE\"\u003e\u003cimg src=\"https://img.shields.io/github/license/ghostapp-ai/ghost?style=flat-square\u0026color=green\" alt=\"License\" /\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/ghostapp-ai/ghost/actions/workflows/ghost.yml\"\u003e\u003cimg src=\"https://img.shields.io/github/actions/workflow/status/ghostapp-ai/ghost/ghost.yml?branch=main\u0026style=flat-square\u0026label=CI\" alt=\"CI\" /\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/ghostapp-ai/ghost/issues\"\u003e\u003cimg src=\"https://img.shields.io/github/issues/ghostapp-ai/ghost?style=flat-square\" alt=\"Issues\" /\u003e\u003c/a\u003e\n  \u003cimg src=\"https://img.shields.io/badge/platform-Windows%20%7C%20macOS%20%7C%20Linux%20%7C%20Android-lightgrey?style=flat-square\" alt=\"Platform\" /\u003e\n  \u003cimg src=\"https://img.shields.io/badge/privacy-100%25%20local-brightgreen?style=flat-square\" alt=\"Privacy\" /\u003e\n\u003c/p\u003e\n\n---\n\nGhost is a private, local-first **Agent OS** for desktop and mobile. It indexes your files, understands your context, connects to thousands of tools via open protocols (MCP, A2A, AG-UI, A2UI, WebMCP), and takes actions on your behalf — all without sending a single byte to the cloud.\n\nThink **Raycast + Semantic Search + Local AI Agent + Universal Protocol Hub** — but private by design.\n\n## Why Ghost?\n\n- **100% Local**: Your data never leaves your machine. Zero telemetry, zero cloud dependencies.\n- **Instant Search**: Hybrid keyword (FTS5) + semantic vector search across all your documents.\n- **Native AI**: In-process embedding inference via Candle — no Ollama or internet required after first model download.\n- **Lightweight**: \u003c10MB installer, \u003c40MB RAM idle, \u003c500ms launch. 70% less RAM than Electron.\n- **Protocol Hub**: MCP Server + Client, A2A multi-agent coordination, AG-UI streaming, A2UI generative UI — the most connected local agent.\n- **Extensible**: Skills.md plugin system + compatible with 10,000+ MCP servers + future Skills Marketplace.\n\n## Features\n\n### Phase 0 — Foundation (**Complete**)\n\n- Tauri v2 desktop shell with React/TypeScript frontend\n- Rust core engine with `thiserror` error handling + `tracing` logging\n- SQLite + sqlite-vec (via FFI auto-extension) for unified storage (documents, vectors, text)\n- FTS5 keyword search + sqlite-vec KNN vector search with RRF hybrid ranking\n- **Native AI inference**: Candle (HuggingFace Rust ML) with all-MiniLM-L6-v2 (384D, ~23MB) — works on any CPU, zero external dependencies\n- **Fallback chain**: Native → Ollama → FTS5-only keyword search\n- Hardware detection: CPU cores, AVX2/NEON SIMD, GPU backend (CUDA/Metal/Vulkan)\n- File watcher (`notify` + `notify-debouncer-mini`) for real-time document indexing\n- Text extraction: PDF (lopdf), DOCX (zip), XLSX (calamine), TXT, Markdown, code files\n- 27 unit tests passing, zero compiler warnings\n\n### Phase 1 — The Search Bar (**Complete**)\n\n- **Spotlight-like floating window**: `Ctrl/Cmd+Space` global shortcut, decorationless, always-on-top, transparent\n- Auto-hide on focus loss, Escape to dismiss, draggable title region\n- Dark ghost-themed UI with Tailwind CSS v4\n- Debounced search input (150ms) with loading skeletons\n- Virtualized results list (`@tanstack/react-virtual`) with file type icons\n- **Open files**: Enter key or double-click opens files with system default app\n- Keyboard navigation: arrow keys, Enter to open, Esc to dismiss\n- Settings panel: watched directory management, persistent settings (JSON)\n- **Zero-config auto-indexing**: auto-discovers user directories (Documents, Desktop, Downloads, Pictures) on first launch — like Spotlight/Alfred\n- Cross-platform directory detection via `dirs` crate (XDG, Windows Known Folders, macOS standard paths)\n- Auto-start file watcher on launch with saved directories\n- **Reliable window dragging**: programmatic `startDragging()` fallback for Linux/Wayland compatibility\n- 50+ source code extensions indexed (rs, py, js, ts, go, java, cpp, etc.)\n- Status bar: document count, AI engine status, vector search status, chat model status\n- Cross-platform CI/CD: Windows, macOS (ARM64 + Intel), Linux installers\n- 27 tests, zero warnings, ~203KB JS bundle\n\n### Native Chat Engine (**Complete**)\n\n- **Hardware-aware auto-selection**: Detects CPU cores, RAM, GPU at startup → recommends largest fitting model\n- **Model registry**: Qwen2.5-Instruct GGUF family (0.5B/1.5B/3B/7B) with Q4_K_M quantization\n- **Zero-config flow**: detect hardware → pick model → auto-download from HuggingFace Hub → background load\n- **Device selection**: CPU (default), CUDA (`--features cuda`), Metal (`--features metal`)\n- **Chat UI**: Tab-based interface (Search ↔ Chat), message bubbles, model status, loading states\n- **Debug panel**: Collapsible log viewer with pause/resume and color-coded levels (Ctrl+D)\n- **Fallback chain**: Native Candle GGUF → Ollama HTTP API → offline\n- **Settings**: model, device, max_tokens, temperature — all configurable, all with sensible defaults\n- **RAM detection**: Linux (/proc/meminfo), macOS (sysctl), Windows (PowerShell Get-CimInstance)\n\n### First-Launch Experience \u0026 Installer (**Complete**)\n\n- **Onboarding wizard**: Multi-step setup shown only on first launch\n  - Welcome screen with Ghost branding\n  - Hardware auto-detection (CPU, RAM, GPU, SIMD)\n  - Recommended model display with specs (size, RAM requirements, parameters)\n  - One-click model download with real-time progress bar\n  - Setup complete summary with keyboard shortcut reminder\n  - Skip option for power users who want to configure later\n- **System tray icon**: Background presence with Show/Quit menu, left-click focus\n- **Professional installer configuration**:\n  - Windows: NSIS with language selector, custom icons, WebView2 silent bootstrap\n  - macOS: DMG with custom layout, minimum macOS 10.15\n  - Linux: DEB (Debian/Ubuntu), RPM (Fedora/RHEL), AppImage (universal)\n- **Filesystem browser**: Navigate directories visually from Settings\n- **OneDrive-aware indexing**: Detects cloud placeholders, indexes metadata only\n- **Zero-config**: Auto-discovers Documents, Desktop, Downloads, Pictures on first launch\n- **Settings persistence**: `setup_complete`, `launch_on_startup`, all chat preferences with serde defaults\n\n### Phase 1.7 — Multiplatform (**Complete**)\n\n- **Android APK**: Full Tauri v2 mobile build (39MB APK, 16MB AAB for aarch64)\n- **Conditional compilation**: `#[cfg(desktop)]` / `#[cfg(mobile)]` for platform-specific code\n- **TLS migration**: 100% rustls — zero OpenSSL, Android NDK cross-compilation safe\n- **Desktop-only gating**: llama-cpp-2, file watcher, system tray, global shortcuts, MCP stdio\n- **Responsive frontend**: all components adapted with `isMobile` prop, 44px+ touch targets, safe areas\n- **Platform detection**: `usePlatform()` hook for runtime UI adaptation\n- **iOS ready**: backend + frontend fully adapted, scaffold requires macOS\n\n### Phase 1.5 — The Protocol Bridge *(In Progress)*\n\n- **MCP Server**: ✅ Expose Ghost tools (search, index, stats) to Claude, Cursor, VS Code via `rmcp` + HTTP streamable transport\n- **MCP Client**: ✅ Connect to external MCP servers (filesystem, GitHub, databases, 10,000+) via stdio + HTTP\n- **AG-UI Runtime**: ✅ Bidirectional agent↔user streaming (~16 event types) — event bus, SSE endpoint, `useAgui` React hook, streaming chat\n- **A2UI Renderer**: Generative UI from JSON schemas — forms, tables, charts rendered natively in React\n- **MCP Apps**: Interactive tool UIs in sandboxed iframes within Ghost conversations\n- **Skills.md**: OpenClaw-inspired plugin format — plain Markdown skill definitions\n\n### Phase 2 — The Agent OS\n\n- **A2A Protocol**: Agent-to-Agent coordination — Ghost delegates to specialized agents\n- **Tool Calling Engine**: Qwen2.5-7B selects + invokes MCP tools from schemas\n- **OS Integration**: UI Automation (Windows), accessibility APIs, clipboard history, browser history\n- **Micro-agents**: Background agents (file organizer, meeting summarizer, email drafter)\n- Premium features: sync, encryption, advanced models, unlimited automations\n\n### Phase 2.5 — The Web Agent\n\n- **WebMCP**: Read tool contracts from websites via W3C `navigator.modelContext` API\n- **Browser Extension**: Bridge between Ghost desktop agent and web-based tool contracts\n- Structured web interactions without scraping\n\n### Phase 3 — The Platform\n\n- Skills Marketplace: third-party skill distribution and monetization\n- Integrations: Obsidian, VS Code, Slack, Notion, browsers\n- Multi-agent orchestration: A2A task delegation between local agents\n- B2B/Teams: shared vaults, SSO, audit trails, compliance\n\n## Architecture\n\nGhost uses a 6-layer **Agent OS** architecture where each layer is independently replaceable:\n\n```text\n┌──────────────────────────────────────────────────────┐\n│              Frontend (React/TypeScript)               │\n│  Omnibox │ Results │ Chat │ A2UI Renderer │ Settings  │\n├──────────────────────────────────────────────────────┤\n│         AG-UI Runtime (Agent ↔ User Streaming)        │\n│  ~16 event types │ Human-in-the-loop │ State sync      │\n├──────────────────────────────────────────────────────┤\n│              Tauri v2 IPC Bridge                       │\n├──────────────────────────────────────────────────────┤\n│              Protocol Hub (Rust — rmcp + custom)       │\n│  MCP Server │ MCP Client │ A2A │ WebMCP │ Skills       │\n├──────────────────────────────────────────────────────┤\n│              Core Engine (Rust)                        │\n│  File Watcher │ Text Extractor │ Embedding Engine      │\n│  Vector DB │ OS Automation │ Micro-agents              │\n├──────────────────────────────────────────────────────┤\n│              AI Layer (Local — Zero Dependencies)      │\n│  Native: Candle + all-MiniLM-L6-v2 (384D embeddings)  │\n│  Fallback: Ollama + nomic-embed-text (768D)            │\n│  Chat: Qwen2.5-Instruct GGUF (0.5B–7B, tool calling)  │\n└──────────────────────────────────────────────────────┘\n```\n\n### Hybrid Trigger System\n\nGhost uses a two-speed architecture to feel instant without burning CPU:\n\n| Layer            | When      | Speed       | Resource Usage   |\n| ---------------- | --------- | ----------- | ---------------- |\n| **Fast Layer**   | Always    | \u003c10ms       | 0% GPU, \u003c1% CPU  |\n| **Smart Layer**  | On demand | 200-2000ms  | Activates Ollama |\n\nThe Fast Layer uses OS accessibility APIs and FTS5 keyword search. The Smart Layer activates only when the user asks a natural language question, requests an action, or a new file needs indexing.\n\n### Tech Stack\n\n| Component          | Technology                       | Why                                                |\n| ------------------ | -------------------------------- | -------------------------------------------------- |\n| Shell/UI           | Tauri v2 + React/TypeScript      | \u003c10MB installer, native performance                |\n| Database           | SQLite + sqlite-vec + FTS5       | Single .db file, vectors + text + metadata         |\n| Native Embeddings  | Candle + all-MiniLM-L6-v2        | 384D, ~23MB, in-process, no external deps          |\n| Fallback Embeddings| Ollama + nomic-embed-text        | 768D, optional, higher quality for large models    |\n| LLM / Chat         | Candle GGUF + Qwen2.5-Instruct   | Native inference, tool calling, 0.5B–7B tiers      |\n| MCP Protocol       | rmcp (official Rust SDK)         | Server + Client, `#[tool]` macro, 10,000+ servers  |\n| Agent Interaction  | AG-UI + A2UI                     | Real-time streaming + generative UI from JSON       |\n| Agent Coordination | A2A (Google)                     | Multi-agent task delegation, Agent Cards            |\n| Web Agent          | WebMCP (W3C)                     | Browser tool contracts, no scraping                 |\n| File Watcher       | notify (Rust crate)              | Cross-platform, async, \u003c1% CPU                     |\n| Text Extraction    | lopdf + zip + calamine           | Pure Rust, no external dependencies                |\n| Encryption         | ChaCha20-Poly1305 (age crate)    | Modern, audited (Pro)                              |\n\n## Getting Started\n\n### Download \u0026 Install\n\nDownload the latest release for your platform from [**GitHub Releases**](https://github.com/ghostapp-ai/ghost/releases/latest):\n\n| Platform | File | Notes |\n|----------|------|-------|\n| **Windows** (64-bit) | `ghost_x.x.x_x64-setup.exe` | NSIS installer, no admin required |\n| **macOS** (Apple Silicon) | `ghost_x.x.x_aarch64.dmg` | M1/M2/M3/M4 Macs |\n| **macOS** (Intel) | `ghost_x.x.x_x64.dmg` | Intel-based Macs |\n| **Linux** (64-bit) | `ghost_x.x.x_amd64.deb` | Debian/Ubuntu |\n| **Linux** (64-bit) | `ghost_x.x.x_amd64.AppImage` | Universal Linux |\n\n| **Android** (ARM64) | `app-universal-release.apk` | Tauri v2 WebView, min SDK 24 |\n\n\u003e **No external dependencies required.** Ghost ships with native AI inference — no Ollama, no GPU, no internet needed after installation.\n\n### Build from Source\n\n#### Prerequisites\n\n- [Rust](https://rustup.rs/) (latest stable)\n- [Node.js](https://nodejs.org/) \u003e= 18 or [Bun](https://bun.sh/) \u003e= 1.0\n- [Ollama](https://ollama.com/) (optional — Ghost uses native AI by default)\n- Platform-specific Tauri v2 dependencies ([see guide](https://v2.tauri.app/start/prerequisites/))\n\n### Installation\n\n```bash\n# Clone the repo\ngit clone https://github.com/ghostapp-ai/ghost.git\ncd ghost\n\n# Install frontend dependencies\nbun install\n\n# (Optional) Pull Ollama model for higher-quality 768D embeddings\n# ollama pull nomic-embed-text\n\n# Run in development mode (native AI model downloads on first run ~23MB)\nbun run tauri dev\n```\n\n### Build for Production\n\n```bash\n# Desktop\nbun run tauri build\n\n# Android (requires Android SDK + NDK 27+)\nbun run tauri android build --target aarch64\n```\n\nThe desktop installer will be generated in `src-tauri/target/release/bundle/`.\nThe Android APK will be in `src-tauri/gen/android/app/build/outputs/apk/`.\n\n## Project Structure\n\n```\nghost/\n├── src/                    # Frontend (React/TypeScript)\n│   ├── components/         # UI components (Onboarding, GhostInput, ResultsList, Settings, StatusBar)\n│   ├── hooks/              # Custom React hooks (useSearch, useHotkey, usePlatform)\n│   ├── lib/                # Tauri IPC wrappers + TypeScript types + mode detection\n│   ├── styles/             # Global CSS (Tailwind v4 theme, safe areas, touch targets)\n│   └── App.tsx             # Root component (onboarding → main UI routing, platform-aware)\n├── src-tauri/              # Backend (Rust)\n│   ├── src/\n│   │   ├── lib.rs          # Tauri commands: search, index, watcher, settings, platform info\n│   │   ├── main.rs         # Entry point\n│   │   ├── error.rs        # Error types (thiserror)\n│   │   ├── settings.rs     # Persistent settings (JSON)\n│   │   ├── chat/           # Chat engine: native Candle GGUF (desktop) + Ollama fallback\n│   │   ├── indexer/        # File watcher (desktop) + text extraction + chunking\n│   │   ├── db/             # SQLite + sqlite-vec + FTS5 (cross-platform FFI types)\n│   │   ├── embeddings/     # Native Candle + Ollama engines + hardware detection\n│   │   ├── search/         # Hybrid search engine + RRF ranking\n│   │   └── protocols/      # MCP server/client, AG-UI, A2A, A2UI, WebMCP\n│   ├── gen/android/        # Generated Android Gradle project (Tauri v2)\n│   ├── capabilities/       # Platform-split permissions (default, desktop, mobile)\n│   ├── Cargo.toml          # Rust deps (target-specific for desktop/mobile)\n│   └── tauri.conf.json     # Tauri configuration + bundler config\n├── branding/               # Brand assets (SVGs, PNGs, social, scripts)\n├── ROADMAP.md              # Detailed development roadmap\n├── CLAUDE.md               # Agent instructions for AI-assisted development\n├── CONTRIBUTING.md         # Contribution guidelines\n├── SECURITY.md             # Security policy \u0026 vulnerability disclosure\n└── package.json            # Frontend dependencies\n```\n\n## Roadmap\n\nSee [ROADMAP.md](ROADMAP.md) for the detailed development plan with phases, milestones, and technical deliverables.\n\n## Privacy \u0026 Security\n\n- **Zero telemetry**: Ghost collects no usage data, no analytics, no crash reports.\n- **Local-only processing**: All AI inference runs on your machine — native Candle engine or optional Ollama.\n- **Single file database**: Your entire vault is one `.db` file you control.\n- **Optional encryption**: ChaCha20-Poly1305 for vault encryption when sync is enabled (Phase 2).\n- **Open source core**: The engine is fully auditable.\n\n## Contributing\n\nGhost is currently in early development. Contributions are welcome!\n\nPlease read [CONTRIBUTING.md](CONTRIBUTING.md) for development setup, coding conventions, and PR guidelines.\n\nFor security vulnerabilities, see [SECURITY.md](SECURITY.md).\n\n## License\n\nThis project is licensed under the MIT License — see the [LICENSE](LICENSE) file for details.\n\n## Acknowledgments\n\n- [Tauri](https://tauri.app/) — Desktop framework\n- [Ollama](https://ollama.com/) — Local LLM runtime\n- [sqlite-vec](https://github.com/asg017/sqlite-vec) — Vector search for SQLite\n- [Candle](https://github.com/huggingface/candle) — Rust ML framework for native AI inference\n- [rmcp](https://crates.io/crates/rmcp) — Official Rust MCP SDK\n- [MCP](https://modelcontextprotocol.io/) — Model Context Protocol (Linux Foundation / AAIF)\n- [A2A](https://google.github.io/A2A) — Agent-to-Agent Protocol (Google / Linux Foundation)\n- [AG-UI](https://github.com/CopilotKit/ag-ui) — Agent-User Interaction Protocol (CopilotKit)\n- [OpenClaw](https://github.com/nicepkg/OpenClaw) — Model-agnostic agent infrastructure\n\n---\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003eYour data. Your machine. Your ghost.\u003c/strong\u003e\n\u003c/p\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fghostapp-ai%2Fghost","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fghostapp-ai%2Fghost","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fghostapp-ai%2Fghost/lists"}