docs: add OpenSIN-AI agent comparison guide

.vitepress/config.js

@@ -33,6 +33,7 @@ export default defineConfig({
           { text: 'OpenSIN-AI Code (Python)', link: '/docs/guide/opensin-ai-code' },
           { text: 'Rust Engine', link: '/docs/guide/opensin-code-rust-engine' },
           { text: 'OpenSIN-AI Platform', link: '/docs/guide/opensin-ai-platform' },
+          { text: 'Agent Features vs Competitors', link: '/docs/guide/opensin-ai-agent-features' },
         ],
       },
       {

docs/guide/opensin-ai-agent-features.md

@@ -0,0 +1,133 @@
+# OpenSIN-AI Agent: Feature Comparison & Roadmap
+
+> How OpenSIN-AI Agent compares against OpenClaw, Google Gemini Agent, and Claude Code — and our plan to surpass them all.
+
+---
+
+## Feature Comparison Matrix
+
+| Feature | OpenSIN-AI | OpenClaw | Gemini Agent | Claude Code |
+|---------|:---------:|:--------:|:------------:|:-----------:|
+| **Core** | | | | |
+| Open Source | ✅ | ✅ | ❌ | ❌ |
+| Self-Hosted | ✅ | ✅ | ❌ | ❌ |
+| 24/7 Autonomous | ✅ | ✅ | ❌ | ✅ (3 days) |
+| Multi-Model | ✅ | ✅ | ❌ | ❌ |
+| **Intelligence** | | | | |
+| Heartbeat/Cron | ✅ | ✅ | ✅ (limited) | ✅ (/loop) |
+| Persistent Memory | ✅ | ✅ | ❌ | ✅ |
+| Structured Wiki | ✅ | ✅ | ❌ | ❌ |
+| Session Branching | 🔲 | ✅ | ❌ | ❌ |
+| **Communication** | | | | |
+| Telegram | ✅ | ✅ | ❌ | ✅ |
+| WhatsApp | ✅ | ✅ | ❌ | ❌ |
+| Discord | ✅ | ✅ | ❌ | ✅ |
+| Signal | ✅ | ✅ | ❌ | ❌ |
+| iMessage | ✅ | ✅ | ❌ | ❌ |
+| Slack | 🔲 | ✅ | ❌ | ❌ |
+| MS Teams | 🔲 | ❌ | ❌ | ❌ |
+| **Automation** | | | | |
+| Subagent Spawning | ✅ | ✅ | ❌ | ✅ |
+| External Harness (ACP) | ✅ | ✅ | ❌ | ❌ |
+| Webhook Triggers | ✅ | ✅ | ❌ | ❌ |
+| Unified Agent Orchestrator | ✅ | ✅ | ❌ | ❌ |
+| Browser Automation | ✅ | ✅ | ✅ | ✅ |
+| **Workspace** | | | | |
+| Google Workspace | ✅ | ✅ | ✅ | ❌ |
+| Apple Ecosystem | ✅ | ❌ | ❌ | ❌ |
+| Microsoft 365 | 🔲 | ❌ | ❌ | ❌ |
+| **SDK & Dev** | | | | |
+| Agent SDK | ✅ | ❌ | ❌ | ✅ |
+| Plugin System | ✅ | ✅ | ❌ | ❌ |
+| MCP Support | ✅ | ✅ | ❌ | ✅ |
+| **Safety** | | | | |
+| Approval Hooks | ✅ | ✅ | ❌ | ❌ |
+| Audit Logging | 🔲 | ✅ | ❌ | ❌ |
+| Risk Classification | 🔲 | ✅ | ❌ | ❌ |
+| **Advanced** | | | | |
+| Voice Wake | ✅ (Siri) | ✅ | ✅ | ❌ |
+| Canvas/A2UI | 🔲 | ✅ | ❌ | ❌ |
+| Computer Use | ✅ | ✅ | ✅ | ✅ |
+
+**Legend:** ✅ Implemented | 🔲 Planned | ❌ Not available
+
+---
+
+## OpenSIN-AI Unique Advantages
+
+### 1. A2A Fleet Architecture (92+ Workers)
+Unlike competitors that run as a single process, OpenSIN-AI has **92+ specialized agents** — each with its own CLI, MCP server, and A2A endpoint. This means:
+- **Specialization beats generalization** — each agent is an expert in its domain
+- **Horizontal scaling** — add more agents without changing the core
+- **Fault isolation** — one agent crash doesn't take down the system
+
+### 2. n8n Foundation (Free, Indestructible)
+OpenSIN-AI uses n8n on a free OCI VM as its routing backbone:
+- **Zero cost** — Oracle Cloud Always-Free tier
+- **Visual workflows** — drag-and-drop automation
+- **200+ integrations** — n8n's built-in connector library
+- **No vendor lock-in** — self-hosted, full control
+
+### 3. Apple Ecosystem Native (12 Agents)
+No competitor has native Apple automation:
+- FaceTime, iMessage, Notes, Calendar, Reminders, Safari, Shortcuts
+- macOS system control (AppleScript, System Events)
+- iPhone integration (USB, screenshots, shortcuts)
+- Siri voice commands via Shortcuts
+
+### 4. Triple Trigger System
+OpenSIN-AI supports three independent trigger mechanisms:
+1. **Heartbeat** — Periodic agent check-in (like OpenClaw)
+2. **Cron Scheduler** — Scheduled recurring tasks (like Gemini)
+3. **Webhook** — External event triggers (like OpenClaw)
+
+### 5. Multi-Model Freedom
+Unlike Claude Code (Claude only) or Gemini Agent (Gemini only):
+- OpenAI GPT-5.4 (primary, via Antigravity)
+- Google Gemini 3.1 Pro/Flash (via Antigravity)
+- Anthropic Claude Sonnet/Opus (via Antigravity)
+- NVIDIA Qwen 3.5 (via NIM)
+- Any local model (via Ollama)
+
+---
+
+## Implementation Roadmap
+
+### Phase 1: Foundation (Sprint 1-2)
+- Heartbeat system in `opensin-ai-cli`
+- Multi-model router with failover
+- Cron scheduler and unified agent orchestrator
+- Approval hooks and permission gate integration
+- Unified messaging API
+- Extended memory system
+
+### Phase 2: Reach & Control (Sprint 3-4)
+- Channel-based remote control
+- Generic subagent spawning
+- Progress tracking
+- Approval hooks framework
+
+### Phase 3: Orchestration (Sprint 5-6)
+- External harness control (ACP)
+- Session branching & compaction
+- Audit logging
+
+### Phase 4: Differentiation (Sprint 7-8)
+- Canvas/A2UI workspace
+- Voice interaction
+- Extended Google Workspace (Gmail, Calendar, Slides)
+
+---
+
+## Sprint 1 Verified
+
+- `HeartbeatSystem` now handles autonomous check-ins, queue polling, backoff, and graceful shutdown.
+- `FailoverRouter` now selects across OpenAI OCI, Antigravity, and NIM with health-based fallback.
+- `CronScheduler` now parses cron expressions and executes scheduled tasks.
+- `ApprovalHooks` now enforces risk-based gating for safe, write, destructive, network, financial, and auth actions.
+- `AgentOrchestrator` ties the systems together with `AgentLoop`, `LoopMode`, `SmartModelRouter`, and `PermissionEvaluator`.
+- `CLIAgent` now routes through the orchestrator for model selection and approval policy checks.
+
+---
+
+*See also: [OpenSIN-AI Agent Feature Spec](https://github.com/OpenSIN-AI/OpenSIN-overview/blob/main/docs/opensin-ai-agent-feature-spec.md)*