Managed Agents Limited availability
Managed Agents lets you create and run AI agents on the Aiven Platform. Agents can use connected tools to gather information, investigate issues, and perform tasks across Aiven and other systems.
You define what an agent does, choose the AI model it uses, and give it access to the tools it needs. You can interact with an agent in a chat or configure scheduled tasks to run automatically.
Managed Agents is built on open-source agent technology. Aiven manages the infrastructure required to run your agents, so you don't need to deploy or maintain the underlying infrastructure.
Managed Agents is in limited availability. You need access for each project. For more information, see Enable Managed Agents.
Key concepts
Managed Agents uses the following concepts:
| Concept | Description |
|---|---|
| Agent | An AI agent that you create and run on the Aiven Platform. |
| System instructions | Define what the agent does and how it behaves. |
| AI model | The model that processes the agent's instructions and requests. |
| Tools and integrations | Give the agent access to information and external systems. |
| Task prompt | Defines the task the agent performs when it runs. Aiven sends it to the agent on each scheduled run. |
| Schedule | Lets the agent perform recurring tasks automatically. |
| Chat | Lets you give the agent a task or ask follow-up questions. |
How Managed Agents works
A typical Managed Agents workflow is:
- Enable Managed Agents: Request access and enable Managed Agents for your project. See Enable Managed Agents.
- Create an agent: Start from a template, describe a task, or configure the agent manually. See Create an agent.
- Connect tools and integrations: Choose the built-in tools and integrations the agent can use. See Manage integrations.
- Chat or schedule: Send requests in a chat, or run tasks automatically at a specified time or interval. See Chat with an agent and Schedule an agent.
- Manage the agent: Update its system instructions, AI model, and tools. See Manage an agent.
For example, you can use chat to investigate an incident as it happens, or create a daily schedule that asks the agent to summarize service health.
Enable Managed Agents
You need access to Managed Agents for each project.
- In the Aiven Console, open your project.
- Click Agents > Request access.
- After you have access, click Enable agents.
- Optional: To run Managed Agents in a project VPC, click Enable in VPC.
When to use Managed Agents
Use agents for tasks that involve gathering information, analyzing it, and taking actions through connected tools. You can run these tasks on demand or on a schedule.
For example, you can create an agent to:
- Review Aiven for PostgreSQL® logs, summarize the findings, and send an update to Slack.
- Check consumer lag in Aiven for Apache Kafka® and create a Jira issue when the lag requires action.
- Investigate an incident using information from multiple systems and summarize the findings.
- Run recurring operational checks and report the results to your team.
What an agent can do depends on its system instructions, AI model, and available tools and integrations.
Tools and integrations
You choose which tools each agent can use. Agents can use the following:
- Built-in tools: Web Fetch and Web Search.
- Aiven MCP: Access to services in your Aiven project. For more information, see Aiven MCP.
- Other MCP integrations: Access to external systems, such as Slack, GitHub, and Jira.
For more information, see Manage integrations.
Permissions
When you connect Aiven MCP, you grant the agent access to services in the current project. You assign one of the following MCP roles: Read-only, Read-write, or Full access.
Aiven creates a dedicated identity and a scoped token for the agent automatically. Your project permissions limit the MCP role you can assign. For more information, see Agent permissions.
Related pages