NEU HIER? → START-HERE.md
In 60 Sekunden handlungsfähig — für menschliche Entwickler UND AI-Agenten.
Konsolidierung April 2026 (9 → 4 Code-Repos, Wave 3 abgeschlossen): docs/CONSOLIDATION-2026-04.md Kanonische Repo-Zuordnung: docs/CANONICAL-REPOS.md Offene Rationalisierungs-Tickets: docs/FOLLOWUPS.md
The Ultimate Single Source of Truth für die OpenSIN-AI Organisation
📊 205 Repositories (177 live, 28 archived) | 17 Teams | 149 Worker | 4 Templates | 44 Skills | 6 Websites
🔄 Live counts auto-generated by
scripts/generate-master-index.js. See registry/MASTER_INDEX.md for the full list.
Important
SSOT: Die kanonische OpenCode-Konfiguration liegt unter OpenSIN-AI/Infra-SIN-OpenCode-Stack (v2.2.1, 44 Skills, 27 MCPs, 5 Provider).
Nach jeder Änderung an opencode.json MUSS sin-sync ausgeführt werden.
All CI runs on a self-hosted n8n + OCI VM pipeline. GitHub-hosted runners are explicitly disallowed org-wide (billing failures + zero-cost OCI alternative).
- Webhook DNS (public):
https://n8n.delqhi.com/webhook/opensin-ci - GitHub Action:
OpenSIN-AI/sin-github-action(composite, ~2s curl-only) - Internal infra (IP, API keys, systemd unit names): documented in
Infra-SIN-Dev-Setup(private) — see ops-runbook. Never commit internal IPs, hostnames, or API keys to this repo. - Full setup for a new repo:
docs/best-practices/ci-cd-n8n.md(public-safe redacted version) or docs.opensin.ai/best-practices/ci-cd-n8n.
Scope note: this repo indexes the CI contract. The operational runbook, credentials, and infrastructure addresses live in Infra-SIN-Dev-Setup. See BOUNDARIES.md § Rule 5.
| Domain | Repo | Zweck | Sichtbarkeit |
|---|---|---|---|
| opensin.ai | website-opensin.ai | Open-Source Landing Page & Produktvorstellung | 🔒 Privat |
| chat.opensin.ai | OpenSIN-WebApp | User WebApp mit Login — Dashboard, Agenten-Steuerung, Fleet-Management (Vercel ✅ deployed) | 🔒 Privat |
| my.opensin.ai | website-my.opensin.ai | Premium Abo-Modell: User-Login, Cloud-Agenten, Billing, API-Keys | 🔒 Privat |
| docs.opensin.ai | OpenSIN-documentation | Offizielle Benutzer-Dokumentation (VitePress) | 🌍 Public |
| blog.opensin.ai | Biz-SIN-Blog-Posts | OpenSIN-AI Blog & Marketing | 🌍 Public |
| (internal API) | OpenSIN-backend | A2A Fleet Control Plane (Closed Source) — Backend-API, n8n-Routing, Agenten-Orchestrierung — konsumiert von OpenSIN-WebApp |
🔒 Privat |
| Mandate | Status | Description |
|---|---|---|
| Hacker Bypass | ✅ Active | No-Wall policy for technical barriers (5-Layer stack) |
| CI/CD Mandate | ✅ Active | Exclusive use of n8n + OCI for builds |
| Fleet Sync | ✅ Active | Mandatory sin-sync for global config parity |
| Visual Evidence | ✅ Active | Mandatory screenshots/videos in all reports/issues |
| Parallel Exploration | ✅ Active | 5-10 parallel explore + librarian agents for large codebases |
Die Agenten-Modelle werden durch ein mehrstufiges Konfigurationssystem verwaltet:
| Datei | Zweck | Repo |
|---|---|---|
opencode.json |
Haupt-Config — Provider, Modelle, MCPs, sichtbare Agenten | Infra-SIN-OpenCode-Stack |
oh-my-openagent.json |
Subagenten-Modelle — explore, librarian, oracle, etc. | Infra-SIN-OpenCode-Stack |
oh-my-sin.json |
Zentrales A2A Team Register — alle Teams klassifiziert | Infra-SIN-OpenCode-Stack |
my-sin-team-code.json |
Team Coding Agenten + Modelle | Infra-SIN-OpenCode-Stack |
my-sin-team-worker.json |
Team Worker Agenten + Modelle | Infra-SIN-OpenCode-Stack |
my-sin-team-infrastructure.json |
Team Infra Agenten + Modelle | Infra-SIN-OpenCode-Stack |
Legacy path
Delqhi/upgraded-opencode-stackredirects to the canonical repo above. All new links must use the canonical path.
| Subagent | Modell | Fallback-Kette |
|---|---|---|
| explore | nvidia-nim/stepfun-ai/step-3.5-flash |
gemini-3-flash → gpt-5.4 → gemini-3.1-pro → claude-sonnet → qwen |
| librarian | nvidia-nim/stepfun-ai/step-3.5-flash |
gemini-3-flash → gpt-5.4 → gemini-3.1-pro → claude-sonnet → qwen |
Bei grossen Codebases (100k+ Zeilen, 1000+ Dateien) MUESSEN Agenten 5-10 parallele explore + 5-10 librarian-Agenten starten. Ein einzelner Agent liefert nur ~20% Abdeckung → falsche Entscheidungen.
| Repo | Zweck |
|---|---|
| OpenSIN | Free/Open-Source Core Engine — Python (QueryEngine, Hooks, Tools, MCP, Sandbox, Memory, A2A) |
| OpenSIN-Code | Autonomes CLI + SDK (@opensin/sdk) — Agent Loop, Tool System, Model Routing, Memory, Safety, A2A Transport Layer + Rust Engine (71 Dateien, 37.7K Zeilen) |
| opensin-ai-cli | ⚠��� Rationalization pending — Rust coding CLI (70 Dateien, 34.601 Zeilen, 9 Crates). Overlaps with OpenSIN-Code. Do not extend; see docs/FOLLOWUPS.md § R1 |
| opensin-ai-platform | OpenSIN/opensin_agent_platform/. Do not extend; see docs/FOLLOWUPS.md § R2 |
| OpenSIN-backend | Closed Source Backend — A2A Fleet Control Plane, n8n-Routing, Agenten-Orchestrierung, API für OpenSIN-WebApp |
| OpenSIN-WebApp | User WebApp (Next.js, Vercel ✅) — Login, Dashboard, Agenten-Steuerung (gekoppelt an OpenSIN-backend) |
| Team | Worker | Zweck |
|---|---|---|
| Team-SIN-Apple | 12 | macOS/iOS Automation (Mail, Notes, Calendar, FaceTime, Safari, etc.) |
| Team-SIN-Code-Backend | 3 | Server, OracleCloud, Passwordmanager |
| Team-SIN-Code-Core | 4 | Coding-CEO, Code-AI, Code-DataScience, Code-DevOps |
| Team-SIN-Code-CyberSec | 19 | BugBounty, Cloudflare, 16x Security-Spezialisten |
| Team-SIN-Code-Frontend | 11 | Accessibility, App-Shell, Commerce-UI, Design-Systems, etc. |
| Team-SIN-Commerce | 4 | Shop-Finance, Shop-Logistic, TikTok-Shop, Stripe |
| Team-SIN-Community | 4 | Discord, WhatsApp, Telegram, YouTube Community |
| Team-SIN-Forum | 9 | Reddit, HackerNews, StackOverflow, Quora, DevTo, etc. |
| Team-SIN-Google | 3 | Google-Apps, Google-Chat, Opal |
| Team-SIN-Infrastructure | 15 | Authenticator, Terminal, Storage, Supabase, N8N, CI-CD, etc. |
| Team-SIN-Legal | 8 | ClaimWriter, Patents, Damages, Compliance, Contract, etc. |
| Team-SIN-Media-ComfyUI | 3 | ImageGen, VideoGen, Workflow |
| Team-SIN-Media-Music | 6 | Beats, Producer, Singer, Songwriter, Videogen, Community |
| Team-SIN-Messaging | 19 | WhatsApp, Telegram, Signal, Discord, iMessage, etc. |
| Team-SIN-Microsoft | 9 | 365, Teams, Outlook, OneDrive, Excel, Word, PowerPoint, etc. |
| Team-SIN-Research | 1 | Deep-Research |
| Team-SIN-Social | 19 | TikTok, Instagram, X, LinkedIn, Facebook, YouTube, etc. |
| Repo | Zweck |
|---|---|
| MCP-SIN-chrome-extension | Chrome Extension MCP - Browser Automation |
| MCP-SIN-computer-use | Computer Use MCP - Desktop Control |
| MCP-SIN-in-chrome | In-Chrome MCP - Tab/Content Access |
| MCP-SIN-mcp-gateway | MCP Gateway - Central Routing |
| MCP-SIN-memory | Memory MCP - Persistent Storage |
| MCP-SIN-platform-auth | Platform Auth - OAuth/SSO |
| MCP-SIN-usebrowser | UseBrowser - Web Automation |
| Simone-MCP | Semantic Code Intelligence MCP — Symbol-Navigation, Referenzen, strukturelle Edits, hybrides Memory (PFLICHT für alle Agenten). Liegt noch unter Delqhi/ — Transfer in die Org ist als MAN-6 in docs/FOLLOWUPS.md getrackt. |
| Repo | Zweck |
|---|---|
| Template-SIN-Agent | Standard-Blueprint für neue A2A-Agenten (mit A2A Transport Layer) |
| Template-SIN-Team | Team-Manager Blueprint (Delegation, Lexicon RAG, Retry, Telegram) |
| Template-SIN-Worker | Worker-Prozess Blueprint |
| Template-SIN-Agent-Worker | Agent Worker Template |
| Plugin | Dateien | Zweck |
|---|---|---|
| sin-code-review | 6 | Automatisierte Code-Reviews mit Multi-Agent-System |
| sin-commit-commands | 10 | Git-Workflow-Befehle (Commit, Push, PR) |
| sin-feature-dev | 12 | Feature-Entwicklung mit spezialisierten Agenten |
| sin-frontend-design | 6 | UI/UX-Implementierung und Design-Skill |
| sin-security-guidance | 6 | Security-Reminder-Hook |
| sin-hookify | 48 | Regelbasierte Hook-Erstellung (größtes Plugin) |
| sin-explanatory-mode | 8 | Erklärt Implementierungsentscheidungen |
| sin-learning-mode | 8 | Interaktiver Lernmodus |
| sin-model-migration | 10 | Migration zwischen Sin-Modellen |
| sin-agent-sdk-dev | 10 | SDK-Entwicklung für Agenten |
| sin-plugin-dev | 116 | Plugin-Entwicklungs-Toolkit (7 Skills, 100+ Docs) |
| sin-pr-review | 18 | PR-Review-Toolkit (6 Agenten) |
| sin-loop | 16 | Selbst-referenzielle Entwicklungsloops |
| sin-ralph | 7 | Ralph-Loop-Technik |
| Repo | Zweck |
|---|---|
| Plugin-SIN-Biometrics | Governance, Policy Enforcement, Supervisor für OpenCode |
| Plugin-SIN-Swarm | Subagent-Workflow mit tmux, Parallel Fan-Out, MAX-Mode |
| Skill-SIN-Agent-Forge | OpenCode Skill: A2A-Agenten erschaffen |
| Skill-SIN-Create-TelegramBot | OpenCode Skill: Telegram-Bot Setup |
| Skill-SIN-Enterprise-Deep-Debug | OpenCode Skill: Strukturiertes Debugging |
GitLab Storage ist tot! GitLab hat unser Konto gesperrt. Ab sofort Box.com (10 GB free) + Google Drive (15 GB free).
| Provider | Storage | Purpose | Link |
|---|---|---|---|
| Box.com Public | 10 GB | Logos, Bilder, Docs — öffentlich | Link |
| Box.com Cache | 10 GB | Logs, Cache, Temp Files | Link |
| Google Drive | 15 GB | User-Daten, Backup | Link |
→ Vollständige Box Storage Doku
Note: As of April 2026,
OpenSIN-onboardinghas been consolidated intoInfra-SIN-Dev-Setup/user-onboarding/. See docs/CANONICAL-REPOS.md for the complete archival map.
| Repo | Zweck |
|---|---|
| CLI-SIN-TelegramBot | Fleet-weites Telegram CLI + MCP |
| CLI-SIN-Repo-Sync | Repository Synchronisation |
| Core-SIN-Control-Plane | Doctor/Preflight/Eval Layer |
| Infra-SIN-Docker-Empire | 26-Container Docker Infrastruktur |
| Infra-SIN-Dev-Setup | Development Environment Setup + Box Storage Guide |
| OpenSIN-backend | 🔒 Proprietär: Fleet Control Plane, Stripe, OAuth, Premium-API (ehemals cloud-backend, seit 2026-04 konsolidiert) |
| Repo | Zweck |
|---|---|
| OpenSIN-overview | SSOT: START-HERE.md, LAUNCH-CHECKLIST.md, CANONICAL-REPOS.md, MASTER_INDEX.md (auto-generated, 205 repos) |
| OpenSIN-documentation | Offizielle Doku (VitePress) — docs.opensin.ai |
| OpenSIN-Code | Core Engine Doku: OpenSIN Code, Rust Engine, Plugins |
| opensin-ai-cli | |
| opensin-ai-platform |
| Repo | Status |
|---|---|
OpenSIN-Storage |
🧊 Archiviert → Nachfolger: A2A-SIN-Storage ✅ |
OpenSIN-Ledger |
🧊 Umbenannt → Biz-SIN-Ledger ✅ |
documentation |
🧊 Archiviert → Nachfolger: OpenSIN-documentation ✅ |
| ~40 weitere Legacy-Repos | 🧊 Alle archiviert, Code erhalten |
- Lies die Regeln: → AGENTS.md und START-HERE.md
- Verstehe die Architektur: → registry/MASTER_INDEX.md
- Wähle dein Team: → Siehe Team-Orchestratoren oben
- Erstelle deinen Agenten: → Nutze
Template-SIN-Agentals Blueprint - Sync die Flotte: → Führe
sin-syncnach jeder Config-Änderung aus - Aktiviere Simone MCP + PCPM: → Simone MCP für semantische Code-Navigation, PCPM für Cross-Session-Memory — beide sind Pflicht für jeden Agenten
Ziel: OpenSIN-AI Agent muss alle Wettbewerber übertreffen.
| Feature | OpenSIN-AI | OpenClaw | Gemini Agent | Claude Code |
|---|---|---|---|---|
| Open Source + Self-Hosted | ✅ | ✅ | ❌ | ❌ |
| 24/7 Autonomous (Heartbeat+Cron) | ✅ | ✅ | ❌ | ✅ (3d) |
| Multi-Model + Failover | ✅ | ✅ | ❌ | ❌ |
| Multi-Channel (19+ Platformen) | ✅ | ✅ | ❌ | ✅ (2) |
| Subagent Fleet (92+ Workers) | ✅ | ✅ | ❌ | ✅ (3) |
| External Harness (ACP) | ✅ | ✅ | ❌ | ❌ |
| Google Workspace Deep | ✅ | ✅ | ✅ | ❌ |
| Apple Ecosystem (12 Agents) | ✅ | ❌ | ❌ | ❌ |
| Agent SDK | ✅ | ❌ | ❌ | ✅ |
| Persistent Memory (Wiki+Journals) | ✅ | ✅ | ❌ | ✅ |
| Webhook + Cron + Heartbeat | ✅ | ✅ | ❌ | ❌ |
| Unified Agent Orchestrator | ✅ | ✅ | ❌ | ❌ |
| Approval Hooks | ✅ | ✅ | ❌ | ❌ |
| Session Branching | 🔲 | ✅ | ❌ | ❌ |
| Canvas/A2UI | 🔲 | ✅ | ❌ | ❌ |
| Voice Interaction | ✅ (Siri) | ✅ | ✅ | ❌ |
→ docs/opensin-ai-agent-feature-spec.md
- Sprint 1: Heartbeat, Multi-Model Router, Approval Hooks, Cron Scheduler, Unified Orchestrator
- Sprint 2: Messaging API, Memory Wiki, Remote Control, Subagent Spawning
- Sprint 3-4: ACP Harness, Session Branching, Audit Logging
- Sprint 5-6: Canvas/A2UI, Voice, Google Workspace Deep
HeartbeatSystemkoordiniert autonome Check-ins, Task-Queue-Polling und Graceful Shutdown.FailoverRouterroutet zwischen OpenAI OCI, Antigravity und NIM mit Health-basierter Fallback-Kette.CronSchedulerplant und führt wiederkehrende Aufgaben aus.ApprovalHookserzwingt riskobasierte Freigaben für destruktive, Netzwerk-, Finanz- und Auth-Operationen.AgentOrchestratorverbindet alle vier Systeme mitAgentLoop,LoopMode,SmartModelRouterundPermissionEvaluator.CLIAgentnutzt den Orchestrator jetzt für Modellrouting und Policy-Prüfung.
Zuletzt aktualisiert: 2026-04-09 | OpenSIN-AI CEO Audit
New agents are configured via the /create-a2a-sin-agent skill.
Documentation: How to Configure Agents