Calame
Visual no-code generator that turns any database into multiple scoped MCP servers — one per access group, with PII masking and fail-closed query scoping built in.
README
<p align="center"> <img src="./docs/assets/logo.png" alt="Calame" width="120" /> </p>
<h1 align="center">Calame</h1>
<p align="center"> <strong>Turn any database into an MCP server — visually.</strong> </p>
<p align="center"> <a href="#quick-start">Quick start</a> · <a href="#features">Features</a> · <a href="#how-it-works">How it works</a> · <a href="#license">License</a> </p>
Calame is a self-hosted web app that connects to your PostgreSQL, MySQL, or SQLite database, lets you configure access profiles — per-table permissions, PII masking, row-level scoping — and serves them as MCP (Model Context Protocol) servers with fine-grained auth.
Plug your profile into any MCP client (LLM, Claude Desktop, Cursor, VS Code, ChatGPT Desktop…) or use the built-in chat to query your data in natural language. No code generation, no lock-in — Calame is the control plane.
Quick start
git clone https://github.com/mgasnier95/calame.git
cd calame
pnpm install
pnpm dev
Open http://localhost:4567 — create the admin account, connect a database, create a profile, and you're live.
Or with Docker:
docker compose up
On the first run Calame auto-generates a CALAME_SECRET_KEY used to encrypt
tokens and connection strings, and persists it next to your database
(.calame-secret). If you deploy with Docker, mount a persistent volume on
/data so that file survives restarts — otherwise every restart invalidates
your saved tokens. You can also set CALAME_SECRET_KEY yourself via the
environment to reuse an existing secret.
→ Detailed Quick Start — full walkthrough from install to your first MCP client query (~15 min).
Features
Database connectors
- PostgreSQL · MySQL · SQLite — schema introspection, relations, sample data
- Read-only by design (
SET TRANSACTION READ ONLY), parameterized queries only - Optional SSH tunneling for remote databases
Access profiles
- Pick tables & columns to expose per profile
- PII detection & masking (auto + custom rules, global or per-column)
- Row-level data scoping (e.g. restrict a profile to
client_email = X) - Write queue with approval workflow for mutating queries
Auth — per profile
- Open, Bearer token, password, OIDC SSO, OAuth 2.1 (Google, Microsoft, GitHub…), or external validation URL
- Per-user tokens with revocation
- MCP OAuth 2.1 Dynamic Client Registration (Claude Desktop / Cursor / VS Code auto-discover)
- Full audit log with export
Built-in chat
- Query your data in natural language from the UI — no external client needed
- Pluggable LLM providers:
- Anthropic (Claude direct)
- OpenRouter (Claude, GPT, Gemini, Llama, …)
- Custom OpenAI-compatible — self-hosted Ollama, vLLM, LM Studio
Operations
- Email invitations (SMTP), user management, metrics dashboard
- HashiCorp Vault integration for secrets
- Docker + reverse-proxy templates (Caddy, nginx) included
How it works
┌────────────┐ ┌─────────────┐ ┌───────────────────────────────┐
│ Your DB │──▶│ Calame │──▶│ MCP client (Claude Desktop, │
│ (Pg/My/Lt) │ │ (profiles, │ │ Cursor, VS Code, ChatGPT, │
│ │ │ auth, PII) │ │ built-in chat, LLM…) │
└────────────┘ └─────────────┘ └───────────────────────────────┘
- Connect a database (Calame introspects the schema).
- Create a profile — pick tables, mask PII columns, pick auth mode.
- Start the profile. Its MCP endpoint is
http://localhost:4567/mcp/<profile>. - Point your MCP client at it, or open the built-in chat.
Feedback
Something broken? Have a use case we missed? Open an issue or start a discussion.
License
Calame is dual-licensed:
- Apache 2.0 — root,
packages/*,scripts/, and everything else outsideee/. SeeLICENSE. - Business Source License 1.1 (BUSL-1.1) — the entire
ee/directory (currentlyee/sso, the SSO/OIDC implementation). Seeee/LICENSE.BUSLandee/README.md.
In short: you can self-host, fork, and modify Calame freely; the BUSL on ee/* only restricts repackaging it as a paid competing product. Each BUSL-licensed version automatically converts to Apache 2.0 four years after its publication.
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。