什么是 io.github.ozten/cantrip?
为 Cantrip 提供的 MCP 服务器,Cantrip 是面向独立创业者的 AI 驱动 GTM 引擎。
README
mcp-server-cantrip
Cantrip helps technical founders find their first customers with structured GTM workflows.
- Nail your ICP — define ideal customer profiles, buyer pains, and value props
- Research competitors — map the landscape and find where you can win
- Get next actions — gap analysis tells you exactly what to work on, then does it for you
Works with Claude Code / Claude Desktop / Cursor — any MCP-compatible agent.
Quick start
- Get your API key at cantrip.ai (settings)
- Add the server:
claude mcp add cantrip -e CANTRIP_API_KEY=your-api-key -- npx -y mcp-server-cantrip
- Try this prompt:
"Initialize a Cantrip project for my product: [describe yours in one sentence]"
Configuration
The Quick start command above covers Claude Code. Verify it worked with claude mcp list or the /mcp command inside a session.
Claude Desktop / Cursor / manual JSON
All MCP clients use the same JSON block. Add it to the appropriate config file:
| Client | Config file |
|---|---|
| Claude Code (manual) | ~/.claude.json |
| Claude Code (project-scoped) | .mcp.json in project root |
| Claude Desktop (macOS) | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Claude Desktop (Windows) | %APPDATA%\Claude\claude_desktop_config.json |
| Cursor | .cursor/mcp.json (project) or ~/.cursor/mcp.json (global) |
{
"mcpServers": {
"cantrip": {
"command": "npx",
"args": ["-y", "mcp-server-cantrip"],
"env": {
"CANTRIP_API_KEY": "your-api-key"
}
}
}
}
Note: The
-yflag is required — without it,npxprompts for confirmation which hangs over stdio.
Windows (not WSL): Use
"command": "cmd"with"args": ["/c", "npx", "-y", "mcp-server-cantrip"].
Project-scoped with teams: Use
"${CANTRIP_API_KEY}"in the env value so each developer sets the key in their shell profile.
Project Context (.cantrip.json)
Each project directory contains a .cantrip.json file that tells Cantrip which project to target:
{
"project": "my-saas"
}
This file is created automatically by cantrip_init (new project) or cantrip_connect (existing project). The agent manages it — you don't need to create it manually.
Multiple projects on the same machine? Each project directory gets its own .cantrip.json. The agent switches context by working in the right directory.
Examples of How to use Cantrip
After connecting Cantrip, try these prompts in order:
- "Initialize a Cantrip project for my product: [one sentence about what you're building]"
- "Find my likely ICP and top 3 buyer pains."
- "Research 5 competitors and tell me where I can win."
- "Give me a one-week GTM plan."
Troubleshooting
Server not found / no tools appear:
- Run
claude mcp list(Claude Code) or/mcpinside a session to check connection status. - Make sure the config is in the right file — Claude Code uses
~/.claude.jsonor.mcp.json, not~/.claude/mcp.json.
Server hangs on startup:
- Ensure you have
-yin the npx args. Without it, npx waits for interactive confirmation that can never arrive over stdio.
"Cannot reach Cantrip API" errors:
- Verify
CANTRIP_API_KEYis set in theenvblock of your MCP config. - Check that
https://api.cantrip.aiis reachable from your network.
Windows "Connection closed" errors:
- Use
"command": "cmd"with"args": ["/c", "npx", "-y", "mcp-server-cantrip"].
Architecture
Agent (Claude, etc.) ── MCP protocol (stdio) ──> mcp-server-cantrip ── HTTPS POST ──> https://api.cantrip.ai
The MCP server is a thin translation layer. It converts MCP tool calls into {command, args, flags} JSON envelopes and POSTs them to the Cantrip API. Zero business logic — identical contract to the CLI and React UI.
Tools (17)
Setup
| Tool | Description |
|---|---|
cantrip_connect | Connect workspace to a project (reads/writes .cantrip.json) |
cantrip_status | Check if the daemon is running |
cantrip_init | Create a new project and auto-connect |
Core Commands
| Tool | Description |
|---|---|
cantrip_snapshot | Project overview, drill into entity types |
cantrip_next | List gap-analysis opportunities |
cantrip_next_prompt | Generate a context-rich LLM prompt for an opportunity |
cantrip_next_run | Spawn a background agent for an opportunity |
cantrip_history | Query the audit trail |
Review
| Tool | Description |
|---|---|
cantrip_review | List items pending review |
cantrip_review_accept | Accept an inferred entity |
cantrip_review_reject | Reject an inferred entity |
cantrip_review_resolve | Resolve an escalation |
cantrip_review_dismiss | Dismiss an escalation |
Entity CRUD (generic)
| Tool | Description |
|---|---|
cantrip_entity_list | List entities of a type |
cantrip_entity_show | Show entity detail |
cantrip_entity_add | Create a new entity |
cantrip_entity_edit | Edit an existing entity |
Supported entity types: icp, pain_point, value_prop, experiment, channel, competitor, contact
License
MIT
常见问题
io.github.ozten/cantrip 是什么?
为 Cantrip 提供的 MCP 服务器,Cantrip 是面向独立创业者的 AI 驱动 GTM 引擎。
相关 Skills
网页构建器
by anthropics
面向复杂 claude.ai HTML artifact 开发,快速初始化 React + Tailwind CSS + shadcn/ui 项目并打包为单文件 HTML,适合需要状态管理、路由或多组件交互的页面。
✎ 在 claude.ai 里做复杂网页 Artifact 很省心,多组件、状态和路由都能顺手搭起来,React、Tailwind 与 shadcn/ui 组合效率高、成品也更精致。
前端设计
by anthropics
面向组件、页面、海报和 Web 应用开发,按鲜明视觉方向生成可直接落地的前端代码与高质感 UI,适合做 landing page、Dashboard 或美化现有界面,避开千篇一律的 AI 审美。
✎ 想把页面做得既能上线又有设计感,就用前端设计:组件到整站都能产出,难得的是能避开千篇一律的 AI 味。
网页应用测试
by anthropics
用 Playwright 为本地 Web 应用编写自动化测试,支持启动开发服务器、校验前端交互、排查 UI 异常、抓取截图与浏览器日志,适合调试动态页面和回归验证。
✎ 借助 Playwright 一站式验证本地 Web 应用前端功能,调 UI 时还能同步查看日志和截图,定位问题更快。
相关 MCP Server
GitHub
编辑精选by GitHub
GitHub 是 MCP 官方参考服务器,让 Claude 直接读写你的代码仓库和 Issues。
✎ 这个参考服务器解决了开发者想让 AI 安全访问 GitHub 数据的问题,适合需要自动化代码审查或 Issue 管理的团队。但注意它只是参考实现,生产环境得自己加固安全。
Context7 文档查询
编辑精选by Context7
Context7 是实时拉取最新文档和代码示例的智能助手,让你告别过时资料。
✎ 它能解决开发者查找文档时信息滞后的问题,特别适合快速上手新库或跟进更新。不过,依赖外部源可能导致偶尔的数据延迟,建议结合官方文档使用。
by tldraw
tldraw 是让 AI 助手直接在无限画布上绘图和协作的 MCP 服务器。
✎ 这解决了 AI 只能输出文本、无法视觉化协作的痛点——想象让 Claude 帮你画流程图或白板讨论。最适合需要快速原型设计或头脑风暴的开发者。不过,目前它只是个基础连接器,你得自己搭建画布应用才能发挥全部潜力。