Aiven MCP Server – Manage Cloud Data Services
The Aiven MCP server lets MCP-compatible assistants create, inspect, and manage Aiven cloud data services such as PostgreSQL and Apache Kafka. It supports a hosted HTTP server with OAuth sign-in and a local npx-based server using an Aiven API token, with read-only mode available to reduce operational risk.
Overview
Aiven MCP is Aiven's Model Context Protocol server for connecting AI assistants to the Aiven platform. It allows MCP-compatible clients such as Claude Code, Cursor, Claude Desktop, VS Code, Gemini CLI, and other MCP clients to work with Aiven projects and managed open source data services from natural-language workflows.
The server is documented as early availability and is designed for operational use cases around Aiven services, including PostgreSQL®, Apache Kafka®, service plans, metrics, logs, and service configuration. It can be run against Aiven's hosted MCP endpoint or started locally with npx.
What the MCP server enables
Through Aiven MCP, an AI agent can list Aiven projects, inspect service details, work with service configuration, and access context around supported managed services. The documentation highlights service creation and management, PostgreSQL, Kafka, plans, metrics, logs, and integrations as supported areas. The available tool set can also be limited to specific service families such as PostgreSQL, Kafka, or integrations so the assistant stays focused on the task.
Read-only mode is an important operating option. When enabled, it restricts the server to non-destructive operations and prevents create, update, and delete actions. Without read-only restrictions, tools may be capable of changing services, databases, topics, or data.
When to use it
Use Aiven MCP when an AI assistant needs live operational context from an Aiven organization. Practical workflows include listing projects, reviewing service configuration before a change, checking logs or metrics, exploring PostgreSQL or Kafka resources, and helping teams understand how existing Aiven services are arranged.
It is also useful for guided cloud data operations, but production environments should start with read-only mode and narrowly scoped tool access until the organization has reviewed the risk of allowing AI-assisted changes.
Connection and authentication
The hosted server uses HTTP at https://mcp.aiven.live/mcp and authenticates with OAuth 2.0 with PKCE. On first connection, the user's browser opens so they can sign in, authorize MCP access, and select the organization. An Aiven organization admin must enable MCP access under Admin settings > Authentication > Allow MCP connection.
The local server uses npx -y mcp-aiven and requires an Aiven API token in AIVEN_TOKEN. AIVEN_READ_ONLY can be set to true or false to control whether destructive operations are allowed.
Key considerations
Aiven warns that MCP tools can perform destructive operations, including creating, modifying, and deleting services, databases, topics, and data. AI agents can misinterpret natural-language prompts, so users should review tool calls before they run, especially for write or delete operations. Follow least privilege for API tokens, rotate tokens regularly, use read-only mode where possible, and review the AI agent provider's data-use terms before enabling the integration.
Supported Transports
streamable_http
URL: https://mcp.aiven.live/mcp?read_only=true
stdio
Command: npx
Args:
-ymcp-aiven
Frequently Asked Questions
- When should an AI agent use the Aiven MCP server?
- Use Aiven MCP when an agent needs live context from Aiven projects and managed data services, such as listing projects, reviewing PostgreSQL or Kafka service configuration, checking logs or metrics, or assisting with service-management workflows.
- What does Aiven MCP add to an AI agent's capabilities?
- It gives the agent MCP tools connected to the Aiven platform rather than relying on static knowledge. The documented capabilities include working with Aiven projects, services, PostgreSQL, Apache Kafka, plans, metrics, logs, service configuration, and integrations.
- What can the AI agent access or manage through Aiven MCP?
- Depending on the selected tools and permissions, the agent can access Aiven projects and supported service areas, and may be able to create, update, or delete Aiven services, databases, topics, and data. Read-only mode restricts the server to non-destructive operations.
- How is authentication configured for Aiven MCP?
- The hosted HTTP server uses OAuth 2.0 with PKCE and opens a browser on first connection for sign-in and authorization. The local stdio server uses an Aiven API token provided through `AIVEN_TOKEN`; organizations should scope tokens to minimum required permissions and rotate them regularly.
- Which transport should be used for Aiven MCP?
- Use the hosted HTTP endpoint when the MCP client supports remote HTTP connections and OAuth sign-in. Use the local `npx -y mcp-aiven` stdio server when you need a local process controlled by an Aiven API token or when your MCP client is configured for command-based servers.