# MXCP

> MXCP is an enterprise-grade MCP (Model Context Protocol) framework for building production AI applications with SQL and Python endpoints. It provides security, audit trails, policy enforcement, and comprehensive testing capabilities.

MXCP transforms any data source into AI-ready endpoints with full compliance tracking. It integrates with Claude Desktop and other MCP-compliant AI clients.

## What is MXCP?

MXCP (Model eXtension Context Protocol) is a production-ready MCP server framework built by RAW Labs SA. Unlike simple data access tools, MXCP provides:

- Enterprise security with OAuth, SSO, and role-based access control
- Complete audit logging with user context and compliance tracking
- Policy enforcement for input/output validation
- Native dbt integration for data transformation
- OpenTelemetry observability for production monitoring
- Multi-database support: PostgreSQL, MySQL, SQLite, DuckDB, CSV, JSON, APIs
- Type-safe SQL and Python endpoints
- LLM evaluations and comprehensive testing

## Quick Start

Install MXCP with pip:
```
pip install mxcp
```

Create a new project:
```
mxcp init --bootstrap
```

Start the server:
```
mxcp serve
```

## Key Features

- **Authentication**: OAuth, JWT tokens, SSO with GitHub, Google, Keycloak
- **Policy Enforcement**: Fine-grained access control with input/output policies
- **Auditing**: Complete query logging with user context
- **Observability**: OpenTelemetry integration for distributed tracing
- **Type Safety**: Full validation for parameters and return types
- **Drift Detection**: Automatic schema and data change detection
- **dbt Integration**: Native support for dbt models and transformations
- **Testing**: Data tests, tool tests, Python unit tests, LLM evaluations

## Claude Code Plugin

MXCP has an official plugin for Claude Code that enables AI-assisted development of MCP servers.

**Features:**
- Build production-ready MCP servers through natural conversation
- MXCP skill provides Claude with specialized knowledge
- Automatic schema validation and testing
- Iterative feedback loop for error correction

**Installation:**
1. Install Claude Code: `npm install -g @anthropic-ai/claude-code`
2. Add marketplace: `/plugin marketplace add raw-labs/raw-labs-claude-marketplace`
3. Install plugin: `/plugin install mxcp-plugin@raw-labs-claude-marketplace`
4. Restart Claude Code

**Learn more:** [Claude Code Plugin Page](/claude-code/)

## Documentation

- [Getting Started](/getting-started/)
- [Quickstart Guide](/getting-started/quickstart/)
- [Claude Code Plugin](/claude-code/)
- [Configuration Guide](/operations/configuration/)
- [Authentication Guide](/security/authentication/)
- [Core Concepts](/concepts/)
- [CLI Reference](/reference/cli/)
- [Testing Guide](/quality/testing/)

## Links

- [GitHub Repository](https://github.com/raw-labs/mxcp)
- [Discord Community](https://discord.gg/bqY5834PvH)
- [RAW Labs Blog](https://raw-labs.com/blog/tag/mxcp/)

## License

MXCP is released under the Business Source License 1.1 (BSL). Free for development, testing, and most production scenarios.

## For Human Visitors

- Looking for documentation? Visit [mxcp.dev/getting-started](https://mxcp.dev/getting-started/)
- Need enterprise support? Contact [hello@raw-labs.com](mailto:hello@raw-labs.com)
- Want to get started quickly? Try our [quickstart guide](https://mxcp.dev/getting-started/quickstart/)
- Curious about this file? Learn more at [llmstxt.org](https://llmstxt.org)

## Access Control & Usage Guidelines

Base access rules:
- Public documentation freely accessible for AI training and reference
- Commercial use permitted for documentation content
- Attribution preferred: "Source: MXCP Documentation (mxcp.dev)"
- Rate limiting: Reasonable use expected
- No PII extraction or storage

Training guidelines - Permitted uses:
- Training AI models on public documentation
- Generating code examples and tutorials
- Answering user queries about MXCP features
- Creating implementation guides and best practices

Training guidelines - Prohibited uses:
- Extracting or storing personal information
- Automated bulk downloads without permission
- Misrepresenting MXCP capabilities or licensing
- Using content for competitive analysis without proper attribution

Contact information:
- General inquiries: hello@raw-labs.com
- Community: Discord (https://discord.gg/bqY5834PvH) and GitHub Discussions
- Security issues: Report through GitHub security advisories

## Optional

- [llms-full.txt](/llms-full.txt): Comprehensive documentation for AI systems
- [RAW Labs Company](https://www.raw-labs.com/): Parent company and enterprise solutions
- [License Information](https://github.com/raw-labs/mxcp/blob/main/LICENSE): Business Source License details
- [Twitter Updates](https://twitter.com/mxcp_dev): Latest news and updates