weavatrix-online
Provides an online MCP overlay for Weavatrix Cloud and Enterprise, enabling sync, advisory refresh, and architecture contract pull with explicit confirmation.
README
Weavatrix Online
Online MCP overlay for Weavatrix Cloud and licensed Weavatrix Enterprise
deployments. It composes the separately versioned, MIT-licensed, offline-first
weavatrix engine in-process and
owns every outbound capability. The offline package keeps its own version
history and its MIT license unchanged.
weavatrix-online starts at version 0.1.0 and tracks Weavatrix Core 0.3.x
through explicit, compatible dependency updates — it never forks or privately
patches the Core graph, parser, LSP or Health implementations.
What Online adds
Online exposes five tools on top of every offline Core tool, all gated behind
the explicit online capability profile. Nothing leaves the machine without a
local preview and an explicit confirmation.
| Tool | Network | Purpose |
|---|---|---|
online_status |
yes | Discover the configured endpoint's capabilities and accepted payload versions. Sends no repository evidence. |
refresh_advisories |
yes | Query OSV for the active repo's exact npm/PyPI/Go/Maven/Gradle/Cargo versions, validate through Core, refresh the offline advisory cache. |
pull_architecture_contract |
yes | Fetch the owner-approved target architecture for the active opaque repository UUID, validate through Core, cache it locally. |
preview_sync |
no | Serialize the exact, bounded, source-free payload and return a short-lived confirmation token. No request is made. |
sync_graph |
yes | Upload only the payload approved by preview_sync; requires dry_run:false and the matching token. |
Missing remote evidence stays UNKNOWN / NOT_CHECKED, never a clean zero. The
packaged agent skill (skill/SKILL.md) tells assistants to run
online_status → preview_sync → sync_graph in that order.
Configuration (sync targets)
Online reaches a destination only when it is explicitly configured through environment variables:
| Variable | Meaning |
|---|---|
WEAVATRIX_SYNC_URL |
Destination sync endpoint, e.g. https://app.weavatrix.com/api/v1/graphs/sync. HTTPS is required except for loopback. |
WEAVATRIX_SYNC_TOKEN |
Bearer credential issued by the endpoint. Never embed credentials in the URL. |
WEAVATRIX_CAPABILITIES_URL |
Optional explicit capabilities document URL. Defaults to /api/v1/capabilities on the sync origin. |
Supported destinations:
- Weavatrix Cloud —
https://app.weavatrix.com, the multi-tenant SaaS workbench. - Weavatrix Enterprise — a customer-hosted deployment under commercial license.
- Local development / self-host — a loopback origin such as
http://127.0.0.1:8787is the one case where plain HTTP is accepted, so the hosted app can be run and exercised locally underwrangler dev.
The connector never grants an endpoint operator access to customer source; tenant and deployment authorization is enforced by the selected endpoint.
Status
Version 0.1.0 is the first source-available Online overlay for Weavatrix 0.3. It owns advisory refresh, local sync preview + exact-payload confirmation, graph sync, target-architecture contract pull, and Cloud/Enterprise endpoint authentication + capability negotiation. It can register its own proprietary MCP tools, packaged skills and local analyzer providers through the Core extension API; those augment the baseline and never replace Core providers.
Development bootstrap
npm install
npm test
node bin/weavatrix-online.mjs C:\path\to\repo
Core improvements land in the MIT weavatrix repository first; Online receives
them through an explicit compatible dependency update.
Licensing
The overlay code uses the permission-required
Weavatrix Online Source License 1.0; commercial terms are in
COMMERCIAL-LICENSE.md. The separately installed
weavatrix dependency remains MIT and is not relicensed here. The repository is
source-available, not open source as a whole; obtain legal advice before relying
on it for a particular commercial arrangement.
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。