[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82240":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":15,"stars30d":16,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":17,"rankGlobal":10,"rankLanguage":10,"license":18,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":19,"hasPages":19,"topics":21,"createdAt":10,"pushedAt":10,"updatedAt":40,"readmeContent":41,"aiSummary":42,"trendingCount":15,"starSnapshotCount":15,"syncStatus":43,"lastSyncTime":44,"discoverSource":45},82240,"GameClaw","linyshdhhcb\u002FGameClaw","linyshdhhcb","Game Enterprise R&D AI Agent Control Plane — 25+ LLM providers, five-layer security, multi-tenant isolation, game-specific tools with hallucination detection | 面向游戏企业研发的 AI Agent 控制平面 — 25+ LLM 供应商、五层安全、多租户隔离、游戏专属工具 + 幻觉检测","",null,"Java",56,3,1,0,4,42.21,"GNU Lesser General Public License v3.0",false,"master",[22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39],"ai-agent","code-generation","control-plane","game-design","game-development","java","llm","mcp","multi-agent","ollama","openapi","plugin-system","postgresql","rbac","skills","spring-ai","spring-boot","spring-modulith","2026-06-12 04:01:37","\u003Cdiv align=\"center\">\n\n# GameClaw\n\n**Game Enterprise R&D AI Agent Control Plane**\n\n[![Java](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJava-25-orange?logo=openjdk)](https:\u002F\u002Fjdk.java.net\u002F25\u002F)\n[![Spring Boot](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSpring%20Boot-4.0.6-brightgreen?logo=springboot)](https:\u002F\u002Fspring.io\u002Fprojects\u002Fspring-boot)\n[![Spring AI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSpring%20AI-2.0.0--M6-6DB33F?logo=spring)](https:\u002F\u002Fspring.io\u002Fprojects\u002Fspring-ai)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-LGPL%20v3-blue)](LICENSE)\n\n**English** | [简体中文](docs\u002Freadme\u002FREADME.zh-CN.md)\n\n[Features](#features) · [Quick Start](#quick-start) · [Architecture](#architecture) · [Configuration](#configuration) · [Contributing](#contributing)\n\n\u003C\u002Fdiv>\n\n---\n\n## What is GameClaw\n\nGameClaw is an AI Agent control plane designed for game development teams. It enables designers, programmers, QA, and ops to collaborate with AI through natural language, handling daily tasks such as game config generation, code writing, data queries, and test automation.\n\n**Core Values**:\n\n- **Game-Specific** — Built-in Unity \u002F Unreal \u002F Godot API index + hallucination detection; AI won't fabricate non-existent APIs\n- **Five-Layer Security** — Network TLS → Access OAuth2 → Application RBAC → Data RLS → Audit logging, enterprise-grade security out of the box\n- **Multi-Tenant Isolation** — PostgreSQL 16 Row-Level Security, strict data isolation across projects and teams\n- **25+ LLM Providers** — Anthropic \u002F OpenAI \u002F DeepSeek \u002F Ollama \u002F Groq \u002F Qwen \u002F Kimi, switch with a single config\n- **Omnichannel Access** — Web \u002F Feishu \u002F Telegram \u002F Discord, the same Agent across all platforms\n- **Plugin Ecosystem** — OpenClaw L3 compatibility + Skills hot-reload + MCP protocol, extensible by third-party developers\n\n---\n\n## Features\n\n### Game Design Tools\n\n| Tool | Description |\n|------|-------------|\n| `generate_monsters` | Natural language description → monster JSON config table |\n| `generate_skills` | Generate skill configs |\n| `generate_items` | Generate item configs |\n| `generate_quests` | Generate quest configs |\n| `generate_growth_curve` | Generate growth curve CSV |\n| `query_engine_api` | Query engine APIs (\"How to load a scene in Unity\" → SceneManager.LoadScene) |\n\n### Programmer Tools\n\n| Tool | Description |\n|------|-------------|\n| `generate_unity_script` | Requirement → C# code + API hallucination check → write to sandbox |\n| `generate_unreal_script` | Same as above, C++ |\n| `generate_godot_script` | Same as above, GDScript |\n\n### Data Analysis Tools\n\n| Tool | Description |\n|------|-------------|\n| `query_data` | Natural language → SQL → dual-layer validation → execute → PII-masked response |\n\n### Governance & Security\n\n| Layer | Capability |\n|-------|------------|\n| Gate 1 | Schema validation (Jackson + Bean Validation) |\n| Gate 2 | Rule engine (10 default rules + YAML custom rules) |\n| RBAC | 5 risk levels x 10 roles, `@RequireRole` \u002F `@RequireRiskLevel` annotations |\n| Quota | Three-tier quotas (user daily \u002F project monthly \u002F global daily budget) |\n| PII | Auto-masking + role-based decryption |\n\n### Plugins & Extensions\n\n| Capability | Description |\n|------------|-------------|\n| OpenClaw L3 Compatible | `@OpenClawPlugin` + `PluginClassLoader` isolation + resource sandbox |\n| Skills Hot-Reload | Auto-reload within 250ms after SKILL.md file changes |\n| MCP Protocol | Spring AI MCP Client + built-in MCP Server |\n| ClawHub Skill Market | install \u002F search \u002F update CLI commands |\n\n---\n\n## Architecture\n\n![GameClaw Architecture](..\u002Fimg\u002Farchitecture.svg)\n\n\n### Project Structure\n\n```\nGameClaw\u002F\n├── base\u002F                    # Core module (20 sub-packages)\n│   ├── agent\u002F               # Agent core + LLM abstraction\n│   ├── channels\u002F            # Channel registration + session management\n│   ├── compat\u002F              # OpenClaw compatibility layer + L3 plugin system\n│   ├── concurrency\u002F         # StructuredTaskScope wrapper\n│   ├── configuration\u002F       # YAML\u002FJSON dual-format config\n│   ├── cost\u002F                # Three-tier quota management\n│   ├── files\u002F               # YAML frontmatter parsing\n│   ├── governance\u002F          # Four-layer gates + rule engine\n│   ├── mcp\u002F                 # MCP connection config\n│   ├── observability\u002F       # Micrometer metrics + audit\n│   ├── persistence\u002F         # Multi-tenant datasource + RLS\n│   ├── project\u002F             # Project management\n│   ├── security\u002F            # RBAC + PII + Prompt injection detection\n│   ├── skills\u002F              # Skills parsing + hot-reload + ClawHub\n│   ├── tasks\u002F               # JobRunr task scheduling\n│   └── tools\u002F               # Game tools + Data tools + Sandbox\n├── providers\u002F               # LLM providers (7 native + 18 OpenAI-compatible)\n├── plugins\u002F                 # Channel plugins (Feishu\u002FTelegram\u002FDiscord\u002FBrave\u002FPlaywright)\n├── mcp-servers\u002Fjava\u002F        # MCP Server (ClickHouse)\n├── app\u002F                     # Spring Boot entry module + CLI\n└── deploy\u002Fdocker\u002F           # Docker deployment (PG16)\n```\n\n### Tech Stack\n\n| Component | Version | Description |\n|-----------|---------|-------------|\n| Java | 25 | Virtual Threads GA \u002F ScopedValue GA \u002F StructuredTaskScope |\n| Spring Boot | 4.0.6 | Virtual threads auto-enabled |\n| Spring AI | 2.0.0-M6 | LLM abstraction layer + MCP Client |\n| Spring Modulith | 2.0.6 | Module boundary enforcement |\n| PostgreSQL | 16 | Row-Level Security multi-tenant isolation |\n| Resilience4j | 2.4.0 | Bulkhead \u002F RateLimiter \u002F CircuitBreaker |\n| JobRunr | 8.6.0 | Background task scheduling |\n| Flyway | — | Database migration |\n| Testcontainers | 1.21.4 | Integration testing |\n\n---\n\n## Quick Start\n\n### Prerequisites\n\n| Dependency | Min Version | Required | Description |\n|------------|-------------|----------|-------------|\n| JDK | 25 | Yes | [Download jdk-25](https:\u002F\u002Fjdk.java.net\u002F25\u002F) |\n| Maven | 3.9 | Yes | Build + Run |\n| LLM API Key | — | Yes | 25+ providers available (Ollama for zero-cost local) |\n| PostgreSQL | 16 | No | Multi-tenant mode only; embedded H2 for single-tenant |\n| Docker | — | No | PG16 multi-tenant mode only |\n\n### 1. Clone & Build\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flinyshdhhcb\u002FGameClaw.git\ncd GameClaw\n\n# Windows\n$env:JAVA_HOME = \"C:\\Program Files\\Java\\jdk-25\"\n\n# macOS \u002F Linux\nexport JAVA_HOME=\u002Fpath\u002Fto\u002Fjdk-25\n\nmvn compile\n```\n\n### 2. Configure LLM\n\n**Option A: Ollama Local Model (Zero Cost)**\n\n```bash\nollama pull qwen3.5:27b\n```\n\nSelect Ollama in the Onboarding wizard after startup. No API Key required.\n\n**Option B: Cloud API Key**\n\nCreate `app\u002Fsrc\u002Fmain\u002Fresources\u002Fapplication.private.yaml` (already .gitignored):\n\n```yaml\nspring.ai.model.chat: deepseek\nspring.ai.deepseek.api-key: sk-xxx\n```\n\n> `spring.ai.model.chat` must be explicitly set; the default value `unknown` does not activate any model.\n\n### 3. Run\n\n```bash\nmvn spring-boot:run -pl app\n```\n\nAfter successful startup, visit:\n- Web UI: http:\u002F\u002Flocalhost:8090\n- Onboarding wizard: http:\u002F\u002Flocalhost:8090\u002Fonboarding\n- Prometheus: http:\u002F\u002Flocalhost:8090\u002Factuator\u002Fprometheus\n- JobRunr: http:\u002F\u002Flocalhost:8091\u002Fdashboard\n\n### 4. Complete Onboarding\n\nOn first launch, visit http:\u002F\u002Flocalhost:8090\u002Fonboarding → Select provider → Enter API Key → Choose model → Assign role → Start chatting\n\n### 5. Multi-Tenant Mode (Optional)\n\n```bash\ncd deploy\u002Fdocker && docker-compose up -d\n# Update application.yaml: spring.datasource.url \u002F flyway.enabled \u002F multi-tenancy.enabled\nmvn spring-boot:run -pl app\n```\n\n### 6. Configure Channels (Optional)\n\n```yaml\n# Feishu\nagent.channels.feishu.app-id: cli_xxx\nagent.channels.feishu.app-secret: xxx\nagent.channels.feishu.verification-token: xxx\n\n# Telegram\nagent.channels.telegram.token: 123456:ABC-xxx\n\n# Discord\nagent.channels.discord.token: your-discord-bot-token\n```\n\n---\n\n## LLM Providers\n\n### Native Providers\n\n| Provider | Default Model | Recommended Use Case |\n|----------|---------------|---------------------|\n| Anthropic | claude-sonnet-4-6 | Production, Claude series |\n| OpenAI | gpt-5.4 | Production, GPT series |\n| Ollama | qwen3.5:27b | Local development, zero-cost experience |\n| Google Gemini | gemini-3-flash-preview | Production, Gemini series |\n| DeepSeek | deepseek-chat | Deep reasoning (R1), game design assistance |\n| Mistral AI | mistral-large-latest | Code generation (Codestral), EU compliance |\n| MiniMax | MiniMax-Text-01 | Chinese dialogue, long context |\n\n### OpenAI-Compatible Providers (18)\n\nBuilt on `OpenAICompatibleProvider` abstract base class, reusing `spring-ai-starter-model-openai` + custom `base-url`:\n\n| Provider | Recommended Use Case |\n|----------|---------------------|\n| Groq | LPU ultra-low-latency inference |\n| xAI Grok | Grok series models |\n| OpenRouter | 300+ model unified gateway |\n| Hugging Face | Open-source model Serverless API |\n| GitHub Copilot | GitHub Models |\n| Qwen | Alibaba Cloud DashScope |\n| Qianfan | Baidu ERNIE series |\n| Moonshot (Kimi) | Long context 128K |\n| StepFun | Multi-modal |\n| Tencent Cloud (Hunyuan) | Tencent Hunyuan |\n| Volcengine | ByteDance Doubao |\n| BytePlus | Overseas ModelArk |\n| Z.AI (Zhipu GLM) | GLM-4.5 code generation |\n| Xiaomi (MiLM) | Lightweight inference |\n| Alibaba Model Studio | Bailian overseas |\n| SenseNova | Multi-modal |\n| Synthetic | Open-source model hosting |\n| SiliconFlow | Open-source model aggregation |\n\n---\n\n## Plugins & Extensions\n\n### Channel Plugins\n\n| Plugin | Description |\n|--------|-------------|\n| **feishu** | Feishu Bot (HMAC signature + card messages + slash commands) |\n| **telegram** | Telegram Bot (Markdown→HTML + whitelist) |\n| **discord** | Discord Bot (@trigger + whitelist) |\n| **brave** | Brave Web Search |\n| **playwright** | Browser automation (navigate\u002Fclick\u002Fscreenshot\u002FJS execution) |\n\n### MCP Server\n\n| Server | Description |\n|--------|-------------|\n| **clickhouse-mcp-server** | ClickHouse data query (SELECT-only + Bearer Token auth) |\n\n### CLI Commands\n\n```bash\ngameclaw skill install \u003Cname>     # Install skill package\ngameclaw skill search \u003Cquery>     # Search skill market\ngameclaw skill update --all       # Update all skills\ngameclaw skill list --installed   # List installed skills\ngameclaw quota check \u003CtenantId>   # Check quota\ngameclaw quota remaining \u003CtenantId> # View remaining quota\n```\n\n### Engine API Index\n\n| Engine | API Count |\n|--------|-----------|\n| Unity | 269 |\n| Unreal | 151 |\n| Godot | 196 |\n\n---\n\n## Configuration\n\n| Config Key | Default | Description |\n|------------|---------|-------------|\n| `server.port` | 8090 | HTTP port |\n| `spring.ai.model.chat` | unknown | Active LLM provider (must be explicitly set) |\n| `gameclaw.llm.adapter` | spring-ai | LLM adapter (spring-ai \u002F langchain4j) |\n| `gameclaw.workspace` | file:.\u002Fworkspace\u002F | Workspace path |\n| `gameclaw.multi-tenancy.enabled` | false | Multi-tenant switch |\n| `gameclaw.security.mode` | dev | Security mode (dev \u002F sso) |\n| `gameclaw.security.rbac.enabled` | true | RBAC switch |\n| `gameclaw.security.outbound.enabled` | true | Outbound whitelist switch |\n| `gameclaw.audit.enabled` | true | Audit logging switch |\n| `spring.threads.virtual.enabled` | true | Virtual threads switch |\n| `spring.flyway.enabled` | false | Flyway migration switch |\n| `gameclaw.quota.enabled` | true | Quota management switch |\n| `gameclaw.quota.user-daily-limit` | 1.0 | User daily budget (CNY) |\n| `gameclaw.quota.project-monthly-limit` | 1000.0 | Project monthly budget (CNY) |\n| `gameclaw.quota.global-daily-limit` | 10000.0 | Global daily budget (CNY) |\n| `gameclaw.llm.model-map.*` | haiku\u002Fsonnet\u002Fopus | Complexity → model mapping |\n| `gameclaw.llm.fallback-map.*` | sonnet\u002Fhaiku\u002Fsonnet | Complexity → fallback model mapping |\n| `gameclaw.skills.polling-interval` | 0 | Skills polling interval (ms, 0=disabled) |\n| `gameclaw.clawhub.enabled` | false | ClawHub skill market switch |\n| `gameclaw.clawhub.registry-url` | https:\u002F\u002Fregistry.clawhub.io | ClawHub registry |\n| `gameclaw.plugins.enabled` | false | OpenClaw L3 plugin system switch |\n| `spring.messages.basename` | i18n\u002Fmessages | i18n message resource path |\n\n---\n\n## Security Model\n\nGameClaw employs a five-layer defense-in-depth architecture:\n\n| Layer | Component | Description |\n|-------|-----------|-------------|\n| L1 Network | TLS 1.3 + OutboundUrlFilter | Transport encryption + outbound whitelist |\n| L2 Access | Spring Security + OAuth2 | Authentication & authorization |\n| L3 Application | RBAC + PromptSanitizer | 5 risk levels x 10 roles + 7 injection detection patterns |\n| L4 Data | RLS + PiiMasking | PostgreSQL row-level security + PII masking |\n| L5 Audit | AuditLogger | Full-chain audit logging |\n\n---\n\n## Contributing\n\nContributions are welcome! Please follow this workflow:\n\n1. Fork this repository\n2. Create a feature branch (`git checkout -b feature\u002Famazing-feature`)\n3. Commit your changes (`git commit -m 'feat: add amazing feature'`)\n4. Push the branch (`git push origin feature\u002Famazing-feature`)\n5. Create a Pull Request\n\n### Development Guidelines\n\n- JDK 25 + Maven\n- Follow Spring Modulith module boundaries\n- Unit tests with AssertJ + JUnit 5\n- Commit messages follow [Conventional Commits](https:\u002F\u002Fwww.conventionalcommits.org\u002F)\n\n---\n\n## License\n\nGNU Lesser General Public License v3.0 — see [LICENSE](..\u002F..\u002FLICENSE)\n\n---\n\n## Author\n\n**linyi**\n\nContact: jingshuihuayue@qq.com\n\n## References\n\nGameClaw's design and implementation were inspired by the following projects:\n\n| Project | Description |\n|---------|-------------|\n| [OpenClaw](https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw) | AI Agent open-source specification and compatibility interface definitions |\n| [GoClaw](https:\u002F\u002Fgithub.com\u002Fnextlevelbuilder\u002Fgoclaw) | Go implementation of AI Agent control plane |\n| [JavaClaw](https:\u002F\u002Fgithub.com\u002Fjobrunr\u002FJavaClaw) | Java implementation of AI Agent framework, GameClaw's migration predecessor |\n","GameClaw 是一个面向游戏企业研发的 AI Agent 控制平面，支持通过自然语言与 AI 协作处理日常任务，如游戏配置生成、代码编写、数据查询和测试自动化。该项目的核心功能包括针对 Unity\u002FUnreal\u002FGodot 引擎内置 API 索引及幻觉检测，确保 AI 不会生成不存在的 API；五层安全保障机制（从网络 TLS 到审计日志），提供企业级安全防护；利用 PostgreSQL 16 行级安全实现严格的多租户数据隔离；集成超过25个大型语言模型供应商，方便切换使用；支持全渠道访问（Web\u002F飞书\u002FTelegram\u002FDiscord）；以及开放插件生态系统，便于第三方开发者扩展功能。GameClaw 特别适合需要高效协作且注重信息安全的游戏开发团队使用。",2,"2026-06-01 03:57:41","CREATED_QUERY"]