IBM i MCP Server

--- title: "Understanding MCP" description: "A practical introduction to the Model Context Protocol (MCP) for IBM i professionals—what it is, why it matters, and how it works." --- **MCP is an open standard that connects AI agents to data sources through a universal protocol.** Instead of building custom integrations for each AI tool, you implement MCP once and it works with Claude, VSCode, Cursor, and any other MCP-compatible client. <Info> **Think of it like HTTP for AI agents**: HTTP is the universal protocol for web browsers. MCP is becoming the universal protocol for AI agents to access external data and tools. </Info> <CardGroup cols={1}> <Card title="Official MCP Documentation" icon="book" href="https://modelcontextprotocol.io"> Comprehensive protocol specification and guides </Card> </CardGroup> --- ## MCP in Three Sentences 1. **MCP is a standardized protocol** that lets AI agents discover and use tools you provide 2. **One server, all clients** - Build once, works with any MCP-compatible AI tool 3. **Security first** - Controlled access, authentication, and audit trails built into the protocol --- ## Why MCP Matters for IBM i **MCP lets AI agents access IBM i data while respecting your existing security model:** - ✅ **SQL runs as the authenticated user** - IBM i object authorities apply - ✅ **Audit trails maintained** - All operations logged with user context - ✅ **Data stays on IBM i** - No ETL pipelines or data copies required - ✅ **Standard tooling** - OpenTelemetry, structured logging, compliance-ready **Real-world use cases:** - Ask AI "What jobs are consuming CPU?" → Queries QSYS2.ACTIVE_JOB_INFO - Ask AI "Show me sales trends" → Runs your existing SQL queries - Ask AI "Check system health" → Executes monitoring queries you define --- ## How MCP Works ```mermaid graph LR A[AI Agent] -->|Request Tool| B[MCP Server] B -->|Execute SQL| C[IBM i] C -->|Results| B B -->|Response| A style A fill:#e3f2fd,stroke:#1976d2,stroke-width:2px,color:#000 style B fill:#fff3e0,stroke:#f57c00,stroke-width:2px,color:#000 style C fill:#e8f5e9,stroke:#388e3c,stroke-width:2px,color:#000 ``` **The Flow:** 1. **AI agent requests a tool** - "I need to query active jobs" 2. **MCP server validates and executes** - Checks auth, runs SQL as authenticated user 3. **IBM i returns data** - SQL results from DB2 for i or QSYS2 services 4. **Agent receives formatted response** - Structured JSON the AI can understand <Note> For detailed protocol specifications, see the [official MCP documentation](https://modelcontextprotocol.io/docs/concepts/architecture). </Note> --- ## MCP Core Concepts MCP defines three main primitives. For IBM i, **tools** are most important: <CardGroup cols={3}> <Card title="Tools" icon="wrench"> **Operations AI agents can execute** For IBM i: SQL queries, stored procedure calls, system commands [Learn more →](https://modelcontextprotocol.io/docs/concepts/tools) </Card> <Card title="Resources" icon="file"> **Static content AI agents can read** For IBM i: Documentation, schemas, configuration files [Learn more →](https://modelcontextprotocol.io/docs/concepts/resources) </Card> <Card title="Prompts" icon="comment"> **Reusable prompt templates** For IBM i: Common query patterns, analysis templates [Learn more →](https://modelcontextprotocol.io/docs/concepts/prompts) </Card> </CardGroup> **Example Tool (YAML):** ```yaml tools: system_status: description: "Get IBM i system performance metrics" statement: "SELECT * FROM qsys2.system_status_info" ``` This becomes a tool AI agents can discover and call automatically. --- ## IBM i MCP Server: What Makes It Different Our implementation adds IBM i-specific features to the standard MCP protocol: - **YAML SQL Tools** - Define database operations in simple YAML (no coding required) - **IBM i Authentication** - Per-user authentication with RSA encryption - **Authority Respect** - SQL runs as the authenticated user, IBM i authorities apply - **Production Features** - OpenTelemetry, structured logging, audit trails <Note> For detailed architecture and implementation details, see [Server Architecture](/concepts/architecture). </Note> --- ## MCP vs. Traditional Integration | Approach | MCP | REST APIs | Direct DB Access | |----------|-----|-----------|------------------| | **Client Support** | Universal | Custom per client | Custom per client | | **Security** | User-level auth | Varies | Service account | | **Data Movement** | Query in place | Often requires ETL | Direct access | | **Audit Trail** | Built-in | Custom | Database logs only | | **Setup Complexity** | Once | Per client | Per client | <Note> **Key Difference:** MCP is a **standard protocol**. Build once, works with any MCP-compatible AI client—today and in the future. </Note> --- ## Next Steps Ready to try MCP with IBM i? Here's your path: <CardGroup cols={2}> <Card title="Quick Start (15 min)" icon="rocket" href="/quickstart"> Install the server and run your first SQL tool </Card> <Card title="Server Architecture" icon="sitemap" href="/concepts/architecture"> Understand how SQL becomes AI tools </Card> <Card title="Build SQL Tools" icon="database" href="/sql-tools/building-tools"> Create custom tools using YAML </Card> <Card title="Connect AI Clients" icon="plug" href="/clients/overview"> Integrate with Claude, VSCode, Cursor, etc. </Card> </CardGroup> --- ## Learn More About MCP <CardGroup cols={2}> <Card title="Official MCP Docs" icon="book" href="https://modelcontextprotocol.io"> Protocol specification and reference </Card> <Card title="MCP GitHub" icon="github" href="https://github.com/modelcontextprotocol"> SDKs, examples, and community resources </Card> </CardGroup> <Note> **The MCP Advantage**: One standard protocol, universal AI client compatibility, enterprise security built-in. Your IBM i data becomes accessible to AI agents while maintaining your security and compliance standards. </Note>