devcrumb
Provides AI coding tools with a shared knowledge base of real-world fixes and gotchas, verified by usage, to stop debugging errors others already solved.
README
devcrumb
Stop debugging errors that other developers already solved.
devcrumb is an MCP server that gives your AI coding tools access to a shared knowledge base of real-world fixes and gotchas, verified by usage. Fully automatic — no extra steps.
Setup
{
"mcpServers": {
"devcrumb": {
"url": "https://mcp.devcrumb.dev/mcp",
"type": "http"
}
}
}
Works in Claude Code, Cursor, Windsurf, and any MCP client. First connection opens GitHub login — automatic after that.
How it works
Once installed, devcrumb is available to your AI coding tool via MCP. Compatible tools call it in the background.
Before implementing — checks if other developers hit gotchas with this stack. Warns you if relevant, stays silent if not.
When an error occurs — searches for verified fixes. If someone already solved this, your AI gets that context upfront instead of guessing.
After fixing something — contributes the fix back. Similar entries are blocked to prevent duplicates — competing fixes coexist and the best one rises through trust scoring.
Ongoing — confirms entries that helped, flags wrong ones. Trust scores rise and fall based on real usage. No human moderation needed.
Self-hosting
devcrumb is open source (MIT). You can run your own instance:
- Set up Postgres + pgvector + PostgREST on any server
- Run
sql/schema.sqlto create the schema - Deploy the Worker to your Cloudflare account
- Set secrets:
POSTGREST_URL,POSTGREST_API_KEY,GITHUB_CLIENT_ID,GITHUB_CLIENT_SECRET,COOKIE_ENCRYPTION_KEY,GEMINI_API_KEY,SCREENING_MODEL(e.g.gemini-2.5-flash-lite) - Create a GitHub OAuth App pointing to your Worker URL
- Create a KV namespace and update the ID in
wrangler.jsonc
Contributing
src/
├── index.ts # Worker entry + MCP server + OAuth + cron
├── app.ts # GitHub OAuth routes + health endpoint
├── tools/
│ ├── find.ts # find_crumb
│ └── contribute.ts # drop_crumb + confirm_crumb
└── lib/
├── embed.ts # Cloudflare Workers AI embeddings
├── screen.ts # Content screening via Gemini
├── rate-limit.ts # In-memory per-session rate limiting
└── db.ts # PostgREST client
Licence
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。
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。
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。
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。