Yasin-MCP
Provides read-only, structured access to Yasin ecosystem information (project registry, documentation, GitHub state, diagnostics) through the MCP protocol, enabling AI agents to query without direct repository access.
README
Yasin-MCP
Standalone AI/Agent-facing MCP (Model Context Protocol) access and integration layer for the Yasin ecosystem.
Mission
Yasin-MCP exposes read-only, structured access to Yasin ecosystem information (project registry, documentation, GitHub repository state, runtime diagnostics) through the MCP protocol, so AI agents and MCP clients can query the ecosystem's state without needing direct, unbounded access to any individual repository or service.
Status
Phase 0, Issue #1 (Repository Bootstrap and Engineering Baseline) only. No MCP server runtime exists yet. No domain adapter (YASIN-DOCS, GitHub, Yasin-Operations) exists yet. This issue only establishes the package structure, tooling, and the safety primitives (error model, deny-by-default policy, secret handling) that everything else will build on.
A note on architecture provenance
The originating task description for this project states that
YASIN-DOCS is "the canonical architecture and boundary" for
Yasin-MCP. Direct verification against the yusi20006-max/YASIN-DOCS
repository (both a code search for "Yasin-MCP"/"MCP" and a direct
check of PROJECT_REGISTRY.yaml and ECOSYSTEM.md) found no
mention of Yasin-MCP anywhere in that repository as of this issue.
This means the architecture implemented here is derived directly from this project's own Issue #1–#10 descriptions (which are detailed and self-consistent), not from a YASIN-DOCS document that does not currently exist. If/when YASIN-DOCS is updated to include Yasin-MCP, that should become the source of truth and this note should be removed or updated to reflect where it now lives.
Architecture boundary
- Yasin-MCP does not replace YASIN-DOCS, Yasin-Core, Yasin-Agent, Yasin-AI, YasinHub, YasinCLI, or Yasin-Operations.
- Phase 1 is strictly read-only: no repository mutation, no deployment, no lifecycle mutation (start/stop/restart), no memory mutation, no arbitrary shell/command execution.
- No private cross-repository imports. Integrations use public APIs, SDKs, contracts, or explicit read-only adapters.
- Existing Yasin projects must not become dependent on Yasin-MCP.
- A deny-by-default policy boundary (
policies/policy.py) rejects any capability whose name matches a forbidden pattern (exec,shell,command,request,arbitrary,filesystem,deploy,delete,start_/stop_/restart_) or that declares itself mutating while Phase 1 is in effect — enforced at construction time, not just at registration.
Package layout
src/yasin_mcp/
server/ MCP server runtime (Issue #4)
protocol/ MCP protocol types/boundary (Issue #2)
capabilities/ capability descriptor + discovery (Issue #2)
tools/ MCP tool implementations (Issue #4+)
resources/ MCP resource implementations (Issue #4+)
adapters/ domain adapters: YASIN-DOCS, GitHub, Yasin-Operations (Issue #5-7)
policies/ deny-by-default policy boundary
errors/ structured error model (McpError + ErrorCategory)
audit/ structured logging, correlation IDs
config/ configuration model, secret handling
version.py package version, EvidenceStatus enum
Evidence model
Responses that report ecosystem state should tag information with one of:
CONFIRMED— directly observed from a live, authoritative sourceTARGET— documented intent/architecture, not verified against a running systemPROPOSED— a suggestion or plan, not yet implemented anywhereUNRESOLVED— could not be determined; must not be presented as fact
Yasin-Operations integration
Yasin-MCP exposes four read-only MCP tools over Yasin-Operations
(yasin_operations_list_services, yasin_operations_service_status,
yasin_operations_health, yasin_operations_diagnostics), via a
subprocess adapter that never imports the yasin_operations package
directly and is registered only when the yasin-operations
executable is available. See
docs/OPERATIONS_INTEGRATION.md
for the full architecture, safety boundary, availability behavior,
and known limitations.
Development
pip install -e ".[dev]"
pytest
ruff check .
ruff format --check .
mypy src
bandit -q -r src
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。