Aiven MCP

Aiven's MCP server lets an AI client manage managed data infrastructure through named tools. Core service tools list projects, create, update and describe services, and report plan, cloud region and status, so an agent can answer what is deployed where and change it. Kafka tools create and delete topics and produce messages to them. PostgreSQL tools read and write through SQL execution and provision connection pooling. Application tools deploy and redeploy containerized applications. A documentation search tool answers product questions from Aiven's own docs on the hosted server, which keeps configuration advice grounded rather than recalled. Because these tools genuinely create, modify and delete services and data, the server ships with guardrails worth configuring before use. A read only mode disables destructive operations, service scoping restricts which tool categories load at all, and a write allowlist permits named actions inside an otherwise read only deployment. Exposure of service credentials is off unless explicitly enabled. The recommended deployment is the hosted endpoint over HTTP with an OAuth flow using PKCE, which an account administrator must enable for the organization before connections are accepted. A local stdio deployment runs through Node.js with an Aiven API token in the environment instead. Compatible clients include Claude Code, Claude Desktop, Cursor, VS Code and Gemini CLI. It is built for developers and platform operators running Aiven data services who want provisioning and troubleshooting in the same place they write code.