airis-mcp-supabase-selfhost
MCP server for self-hosted Supabase with RLS-aware PostgreSQL and PostgREST layers, enabling safe database introspection, SQL queries, and PostgREST access via natural language.
README
Supabase Self-hosted MCP Server
MCP (Model Context Protocol) server for self-hosted Supabase with RLS-aware PostgreSQL + PostgREST layers.
🌟 Part of the AIRIS Ecosystem
This MCP server is designed to work with the AIRIS Suite - accessible via airis-mcp-gateway for token-efficient integration.
The AIRIS Suite
| Component | Purpose | For Who |
|---|---|---|
| airis-agent | 🧠 Intelligence layer for all editors (confidence checks, deep research, self-review) | All developers using Claude Code, Cursor, Windsurf, Codex, Gemini CLI |
| airis-mcp-gateway | 🚪 Unified MCP proxy with 90% token reduction via lazy loading | Claude Code users who want faster startup |
| mindbase | 💾 Local cross-session memory with semantic search | Developers who want persistent conversation history |
| airis-workspace | 🏗️ Docker-first monorepo manager | Teams building monorepos |
| airiscode | 🖥️ Terminal-first autonomous coding agent | CLI-first developers |
MCP Servers (Included via Gateway)
- airis-mcp-supabase-selfhost (this repo) - Self-hosted Supabase MCP with RLS support
- mindbase - Memory search & storage tools (
mindbase_search,mindbase_store)
Recommended: Install via AIRIS MCP Gateway
This MCP server comes pre-configured with AIRIS MCP Gateway. No additional setup required.
# Install the Gateway (includes this server)
brew install agiletec-inc/tap/airis-mcp-gateway
# Start the gateway
airis-mcp-gateway up
# Add to Claude Code
claude mcp add --transport http airis-mcp-gateway http://api.gateway.localhost:9400/api/v1/mcp
Alternative: Standalone Installation
If you need to run this server independently:
git clone https://github.com/agiletec-inc/airis-mcp-supabase-selfhost.git
cd airis-mcp-supabase-selfhost && pnpm install
What you get with the full suite:
- ✅ Confidence-gated workflows (prevents wrong-direction coding)
- ✅ Deep research with evidence synthesis
- ✅ 94% token reduction via repository indexing
- ✅ Cross-session memory across all editors
- ✅ Self-review and post-implementation validation
✨ Features
- 2-Layer Backend: PostgREST (RLS-respected) + Direct PostgreSQL (schema introspection & diagnostics)
- Token-Optimized: Lazy advertise, schema digest, minimal tool descriptions
- Safety-First: READ_ONLY mode by default, DML/DDL/DCL blocked, SQL validation
- Feature Flags: Selective tool activation (database, docs, postgrest, functions, storage)
- RLS Awareness: PostgREST layer respects Row Level Security policies
- Supabase Compatible: Works with official Supabase self-hosted deployments
🏗️ Architecture
┌─────────────────────────────────────────┐
│ LLM (Claude, GPT, etc.) │
└───────────────┬─────────────────────────┘
│ MCP Protocol (JSON-RPC)
┌───────────────▼─────────────────────────┐
│ Supabase Self-hosted MCP Server │
│ ┌─────────────────────────────────┐ │
│ │ Feature Flags & Safety Guards │ │
│ └─────────────────────────────────┘ │
│ │ │ │
│ ┌──────▼──────┐ ┌─────▼──────┐ │
│ │ PostgREST │ │ Direct │ │
│ │ Layer │ │ PG │ │
│ │ (RLS aware) │ │ (read-only)│ │
│ └──────┬──────┘ └─────┬──────┘ │
└─────────┼──────────────────┼───────────┘
│ │
▼ ▼
┌─────────────────────────────┐
│ Supabase Self-hosted │
│ (PostgreSQL + PostgREST) │
└─────────────────────────────┘
🚀 Quick Start
Prerequisites
- Node.js >= 20.0.0
- Self-hosted Supabase instance running (localhost:54321 or remote)
- PostgreSQL read-only user (recommended)
1. Install Dependencies
pnpm install
2. Configure Environment
cp .env.example .env
# Edit .env with your Supabase credentials
Key Configuration:
# PostgreSQL (use read-only role for safety)
PG_DSN=postgres://mcp_ro:password@127.0.0.1:5432/postgres
# PostgREST endpoint
POSTGREST_URL=http://127.0.0.1:54321/rest/v1
# PostgREST JWT (anon key or service_role key)
POSTGREST_JWT=your_anon_key_here
# Safety mode (recommended: true)
READ_ONLY=true
# Feature flags
FEATURES=database,docs,postgrest
3. Create Read-Only PostgreSQL User (Recommended)
-- Connect to your Supabase PostgreSQL as superuser
CREATE ROLE mcp_ro WITH LOGIN PASSWORD 'secure_password';
GRANT pg_read_all_data TO mcp_ro;
GRANT USAGE ON SCHEMA public TO mcp_ro;
GRANT USAGE ON SCHEMA information_schema TO mcp_ro;
4. Start Server
# Development mode (with auto-reload)
pnpm dev
# Production mode
pnpm build
pnpm start
Server starts on http://localhost:3100
🛠️ Available Tools
MVP Tools (Current Implementation)
Tool Prefix: All tools use sbsh_ prefix to avoid collision with official Supabase MCP and PostgreSQL MCP servers.
1. sbsh_introspect_schema
Get token-optimized schema summary for specified schemas.
{
"name": "sbsh_introspect_schema",
"arguments": {
"schemas": ["public"]
}
}
Output: Digest format with table names, column types, index/constraint counts, RLS status.
2. sbsh_execute_sql
Execute SELECT queries or EXPLAIN plans safely.
{
"name": "sbsh_execute_sql",
"arguments": {
"sql": "SELECT * FROM users WHERE created_at > '2024-01-01' LIMIT 10",
"limit": 100
}
}
Safety: In READ_ONLY mode, DML/DDL/DCL are blocked.
3. sbsh_postgrest_get
GET request via PostgREST with RLS respected.
{
"name": "sbsh_postgrest_get",
"arguments": {
"table": "users",
"query": {
"select": "id,email,created_at",
"eq": "status:active",
"order": "created_at.desc",
"limit": 10
}
}
}
Security: Uses provided JWT, respects RLS policies.
4. sbsh_get_table_doc
Get detailed documentation for a specific table.
{
"name": "sbsh_get_table_doc",
"arguments": {
"table": "public.users"
}
}
Output: Columns, constraints, RLS policies, comments.
🔐 Security Features
READ_ONLY Mode (Default)
- Enabled by default for safety
- Blocks all DML/DDL/DCL operations
- Only allows SELECT and EXPLAIN queries
- Recommended for production use
SQL Validation
- Regex-based detection of mutating operations
- Blocks: INSERT, UPDATE, DELETE, TRUNCATE, ALTER, DROP, CREATE, GRANT, REVOKE, etc.
- Allows: SELECT, EXPLAIN
RLS Respect
- PostgREST layer automatically enforces Row Level Security
- Direct PostgreSQL layer uses read-only role (no write permissions)
- JWT-based authentication for user-scoped access
Feature Flags
Control tool availability via FEATURES environment variable:
database: Schema introspection and SQL executiondocs: Table documentationpostgrest: PostgREST API wrapperfunctions: Edge Functions (not yet implemented)storage: Storage API (not yet implemented)
📊 Token Optimization
Lazy Advertise
- Only advertise enabled features
- Minimal tool descriptions (detailed help on demand)
- Feature flags reduce initial token load
Schema Digest
introspect_schemareturns compressed summary- Full details via
docs.get_table_docon demand - Column format:
name:type!(! = NOT NULL)
Result Sampling
- Configurable row limits (max 1000)
- Field metadata included for context
- Truncation warnings
🔧 Development Roadmap
Phase 1: MVP (Current)
- [x] Project setup
- [x] Core tools: introspect_schema, execute_sql, postgrest.get, docs
- [x] READ_ONLY mode
- [x] Feature flags
- [x] Token optimization
Phase 2: Enhanced Safety
- [ ] Dry-run mode for DML operations
- [ ] Audit logging
- [ ] Rate limiting
- [ ] Query cost estimation
Phase 3: Extended Features
- [ ] Edge Functions support
- [ ] Storage API integration
- [ ] Real-time subscriptions
- [ ] Multi-project support
Phase 4: Advanced Optimization
- [ ] Schema caching with invalidation
- [ ] Result pagination
- [ ] Streaming responses
- [ ] Connection pooling optimization
🤝 Integration
With airis-mcp-gateway
This server is designed to work with dynamic MCP gateway patterns:
// Load only when Supabase tools are needed
await gateway.loadServer('airis-mcp-supabase-selfhost', {
url: 'http://localhost:3100/mcp',
features: 'database,docs'
})
// Use tools
await gateway.call('airis-mcp-supabase-selfhost', 'sbsh_introspect_schema', {})
// Unload after use
await gateway.unloadServer('airis-mcp-supabase-selfhost')
With Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"airis-mcp-supabase-selfhost": {
"command": "node",
"args": ["/path/to/dist/server.js"],
"env": {
"PG_DSN": "postgres://mcp_ro:pass@localhost:5432/postgres",
"POSTGREST_URL": "http://localhost:54321/rest/v1",
"POSTGREST_JWT": "your_jwt_here",
"READ_ONLY": "true",
"FEATURES": "database,docs,postgrest"
}
}
}
}
🔗 Related Projects
Explore other tools in the AIRIS ecosystem:
- airis-mcp-gateway - Unified MCP hub with 90% token reduction
- airis-agent - Intelligence layer for AI coding
- mindbase - Local cross-session memory with semantic search
- airis-workspace - Docker-first monorepo manager
- cmd-ime - macOS IME switcher
- neural - Local LLM translation tool
- airiscode - Terminal-first autonomous coding agent
🤝 Contributing
We welcome contributions! Please:
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'feat: add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
📝 License
MIT
🙏 Acknowledgments
- Inspired by Supabase official MCP
- Built on Model Context Protocol
- PostgreSQL introspection patterns from community MCP servers
💖 Support
agiletec is a one-person studio building these tools full-time and open source. If they earn a spot in your workflow, a sponsorship keeps them maintained and independent.
Built with ❤️ by the Agiletec team
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。