@selibiks/bubble-mcp
MCP server for Bubble.io that provides 28 tools for querying data, auditing schemas, seeding test records, and more, with three security layers and configurable modes.
README
@selibiks/bubble-mcp
MCP server for Bubble.io with 28 tools across 3 layers — query data, audit schemas, seed test records, and more.
Quick Start
# Install dependencies
npm install
# Build
npm run build
# Start (requires configuration — see below)
npm start
Configuration
Environment Variables
BUBBLE_APP_URL=https://your-app.bubbleapps.io
BUBBLE_API_TOKEN=your-api-token-here
BUBBLE_MODE=read-only
BUBBLE_ENVIRONMENT=development
BUBBLE_RATE_LIMIT=60
| Variable | Required | Default | Description |
|---|---|---|---|
BUBBLE_APP_URL |
Yes | — | Your Bubble.io app URL |
BUBBLE_API_TOKEN |
Yes | — | API token from Bubble.io Settings > API |
BUBBLE_MODE |
No | read-only |
read-only, read-write, or admin |
BUBBLE_ENVIRONMENT |
No | development |
development or live |
BUBBLE_RATE_LIMIT |
No | 60 |
Max requests per minute |
Config File Alternative
Create bubble.config.json in the project root:
{
"app_url": "https://your-app.bubbleapps.io",
"api_token": "your-api-token-here",
"mode": "read-only",
"environment": "development",
"rate_limit": 60
}
Environment variables take priority over the config file.
Claude Desktop / Claude Code Integration
Add to your Claude configuration:
{
"mcpServers": {
"bubble": {
"command": "node",
"args": ["/path/to/bubble-mcp/dist/index.js"],
"env": {
"BUBBLE_APP_URL": "https://your-app.bubbleapps.io",
"BUBBLE_API_TOKEN": "your-api-token-here",
"BUBBLE_MODE": "read-only",
"BUBBLE_ENVIRONMENT": "development"
}
}
}
}
Or via the CLI:
claude mcp add bubble -- node /path/to/bubble-mcp/dist/index.js
Security Model
Three-layer defence in depth:
| Layer | What it does |
|---|---|
| Server Mode Gate | Filters tools by mode at startup (read-only < read-write < admin) |
| MCP Permissions | Claude prompts user approval before each tool invocation |
| Bubble Privacy Rules | Bubble.io enforces API-level access via the token's permissions |
Input validation: All user-supplied identifiers (dataType, id, sort_field, workflow_name) are validated against safe character patterns before reaching API URLs. File paths (TDD tools) are sandboxed to the project directory.
Token safety: The API token is never exposed in responses or error messages. The bubble_get_environment tool explicitly excludes it.
Mode Reference
| Capability | read-only | read-write | admin |
|---|---|---|---|
| Query data (search, get) | Yes | Yes | Yes |
| Inspect schema / Swagger docs | Yes | Yes | Yes |
| Create / update records | No | Yes | Yes |
| Trigger workflows | No | Yes | Yes |
| Delete records | No | No | Yes |
| Bulk operations | No | No | Yes |
| Seed / cleanup test data | No | No | Yes |
Tool Reference (28 tools)
Core Tools (11)
| Tool | Mode | Description |
|---|---|---|
bubble_get_schema |
read-only | Fetch the full app schema (all types and fields) from /meta |
bubble_search |
read-only | Search records with constraints, sorting, and pagination |
bubble_get |
read-only | Fetch a single record by type and ID |
bubble_create |
read-write | Create a new record |
bubble_update |
read-write | Partially update a record (PATCH) |
bubble_replace |
read-write | Replace a record entirely (PUT) |
bubble_delete |
admin | Permanently delete a record |
bubble_bulk_create |
admin | Create up to 1000 records in one call |
bubble_trigger_workflow |
read-write | Trigger a backend API workflow by name |
bubble_get_environment |
read-only | Return current server config (mode, environment, rate limit) |
bubble_swagger_docs |
read-only | Fetch the Swagger/OpenAPI spec for the connected app |
Compound Tools (7)
| Tool | Mode | Description |
|---|---|---|
bubble_schema_summary |
read-only | Type counts, relationships, and totals (lightweight vs full schema) |
bubble_privacy_audit |
read-only | Scan schema for sensitive fields, PII, and API write exposure |
bubble_find_orphans |
read-only | Find records with broken foreign key references |
bubble_record_validator |
read-only | Sample records and check for empty/null fields |
bubble_search_all |
read-only | Auto-paginate through all records of a type (up to 10k) |
bubble_field_usage |
read-only | Analyse field population rates and identify dead fields |
bubble_compare_environments |
read-only | Diff schemas between development and live environments |
Developer Tools (10)
| Tool | Mode | Description |
|---|---|---|
bubble_health_check |
read-only | Privacy audit + dead field detection with a 0-100 score |
bubble_export_schema |
read-only | Export schema as TDD-format markdown |
bubble_workflow_map |
read-only | List all API workflows and their parameters |
bubble_tdd_validate |
read-only | Validate a TDD markdown file against the live schema |
bubble_migration_plan |
read-only | Generate ordered migration steps from a TDD file |
bubble_wu_estimate |
read-only | Estimate Workload Units for a Bubble operation |
bubble_suggest_indexes |
read-only | Suggest fields that would benefit from indexes |
bubble_option_set_audit |
read-only | Find text fields that should be option sets |
bubble_seed_data |
admin | Seed test data in dependency order (tracked for cleanup) |
bubble_cleanup_test_data |
admin | Delete all previously seeded test data |
Swagger / OpenAPI
The bubble_swagger_docs tool fetches the auto-generated Swagger spec from your Bubble app:
https://your-app.bubbleapps.io/api/1.1/meta/swagger.json
If you get an error, enable Swagger docs in Bubble: Settings > API > Enable Swagger documentation.
Development
npm install # Install dependencies
npm run dev # Development mode (tsx, no build needed)
npm run build # Compile TypeScript to dist/
npm test # Run all tests (179 tests)
npm run test:watch # Tests in watch mode
npm run lint # ESLint
npm run format # Prettier
npm run typecheck # tsc --noEmit
Defaults
| Setting | Default | Why |
|---|---|---|
| Mode | read-only |
Safe — no mutations possible |
| Environment | development |
Safe — won't touch production |
| Rate limit | 60 req/min | Prevents API abuse |
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。