contextcore-mcp
Provides secure, read-only access to ContextCore project context (vision, audience, scope, tech decisions, etc.) for MCP-aware agents like Cursor and Claude Code, with server-side access enforcement.
README
contextcore-mcp
Give Cursor, Claude Code, and any other MCP-aware agent secure, read-only access to your ContextCore project context — vision, audience, scope, tech decisions, rules, skills, risks, and open questions.
The agent gets exactly the context you can see. Access is enforced server-side by ContextCore's org/team/visibility federation — this client holds no authorization logic and can never widen access.
Most people don't need this package. ContextCore's gateway speaks MCP over HTTP, so any editor that supports remote MCP servers can connect with a URL and a token — no install. Use this package only if your client is stdio-only.
What your agent gets
| Tool | What it returns |
|---|---|
list_projects |
Projects you can access (id, name, tagline) |
get_context_core |
A project's full context as one Markdown document |
read_component |
One component (e.g. vision, scope, tech) as Markdown |
search_context |
Keyword matches across pieces, rules, skills, risks, and problems |
The tool list is fetched from the server at startup rather than baked in here, so it stays current without you upgrading the package.
Setup
1. Generate a token
In ContextCore, open Account → Integrations → Generate token and copy it (shown once).
2. Connect
Remote (preferred). Works in Cursor and Claude Code without installing anything:
{
"mcpServers": {
"contextcore": {
"url": "https://contextcore.md/mcp",
"headers": { "Authorization": "Bearer ${env:CONTEXTCORE_TOKEN}" }
}
}
}
Stdio (this package). For clients that don't support remote servers:
{
"mcpServers": {
"contextcore": {
"command": "npx",
"args": ["-y", "contextcore-mcp"],
"env": { "CONTEXTCORE_TOKEN": "YOUR_TOKEN" }
}
}
}
Cursor reads ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project); Claude Code reads .mcp.json in your project root, or:
claude mcp add contextcore --env CONTEXTCORE_TOKEN=YOUR_TOKEN -- npx -y contextcore-mcp
3. Use it
Ask your agent things like "Load the context for my Loop project" or "What's the target audience according to ContextCore?". It will call the tools automatically.
Configuration
| Env var | Required | Default | Description |
|---|---|---|---|
CONTEXTCORE_TOKEN |
yes | — | Personal Access Token from ContextCore |
CONTEXTCORE_API_URL |
no | https://contextcore.md |
ContextCore deployment base URL |
The pre-rename CONTEXTOS_TOKEN and CONTEXTOS_API_URL are still accepted.
Security
- The token maps to a single ContextCore user. Every read runs through the same federation gates as the web app — restricted context you can't see is never returned.
- Only a SHA-256 hash of your token is stored server-side. Revoke any token anytime from Account → Integrations.
- Never commit your token. Reference it via an environment variable in your MCP config where possible.
License
MIT
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。