Any watsonx agent can now tap into live enterprise data streams—discover, inspect, and debug Confluent Cloud in natural language

What's new

Confluent's Managed MCP Server is now available in the watsonx Orchestrate catalog. Any agent you build on Orchestrate Catalog can now discover, inspect, and troubleshoot Confluent Cloud environments through natural-language interactions, including exploring topics and schemas, querying metrics, and diagnosing connector issues—all through a single, governed catalog entry.

Confluent secures the data side, while watsonx Orchestrate governs the agent side, discovers the server for you, and lets you reuse it across every agent and workflow you run.

Background

Confluent joined IBM Software in March 2026. Real-time data streaming is central to IBM's enterprise AI strategy: an agent is only as useful as the data beneath it, and streaming keeps that data current.

The Managed MCP Server provides a standardized, governed way for watsonx Orchestrate agents to interact with Confluent Cloud. By making it available in the watsonx Orchestrate catalog, developers can discover, inspect, and troubleshoot Confluent Cloud resources through a single catalog entry—eliminating the need for custom integrations.

What it does

Picture it as the “central nervous system” for your streaming platform. An agent walks up and asks a question in natural language; the server knows which levers exist behind the desk, translates the request into the right Confluent Cloud API calls, and returns the answer.

Two key aspects make this flow a critical and trustworthy one for enterprise use, and Orchestrate adds a third layer on top:

1. Operational visibility. The server helps agents discover, diagnose, and understand your Confluent Cloud environment through natural-language interactions.
2. Governance-aware. Because it works through Confluent's real APIs, it respects your existing access controls, schemas, and data contracts — not a side channel around them

Listing it in the Orchestrate catalog adds a third layer: Orchestrate controls which agents, teams, and users can invoke the server, making every interaction observable and auditable.

Because Confluent hosts and runs the server, there's nothing for the customer to operate, and because it lives in the Orchestrate catalog, there's nothing for each team to wire up from scratch.

How to get started

1. In watsonx Orchestrate, go to Build agents and tools and select Create agent (or open an existing one).
2. Define your agent's description, behavior, and instructions.
3. On the agent configuration page, select Add tools → Add from Catalog, then filter by MCP Servers.
4. Add the relevant tools:
   • IBM Confluent Cloud Global—for organization-wide operations: discovering environments, debugging connectors, querying metrics.
   • IBM Confluent Cloud Regional—for region-specific operations: reading topics, messages, and schemas.
5. Test with a few prompts, then deploy.

The following user flow shows how a data platform engineer validates an orders pipeline before a flash sale. Instead of navigating across the Confluent CLI, Cloud Console, and documentation, the engineer completes the workflow through the IBM Confluent Cloud Managed MCP Servers in watsonx Orchestrate.

1 - Build the agent and add the IBM Confluent Cloud Managed MCP Server tools

2 - Deploy

3 - Run the readiness check

Common use cases

• Discover available environments and clusters to understand how streaming resources are organized.
• Explore Kafka topics, reviewing configurations, reading sample messages, and understanding the schemas behind event streams.
• Troubleshoot failing or degraded connectors by reviewing status, logs, metrics, and available diagnostic information.
• Investigate performance issues by analyzing throughput, consumer lag, and other operational metrics.
• Understand what data is available before building a new application, integration, or AI agent.
• Inspect and understand the structure and contents of event streams when onboarding to a new project or team.
• Review schemas and subjects to understand the impact of proposed data model changes.
• Create up-to-date documentation for topics, schemas, connectors, and streaming architectures using information retrieved from Confluent Cloud.

These capabilities can also feed into broader workflows — fraud investigation, patient safety operations, supply chain monitoring — wherever agents need visibility into streaming infrastructure.

Sample prompts to try

• "List my Confluent Cloud environments and clusters."
• "What topics exist in cluster lkc-abc123?"
• "What's the throughput for this cluster over the last hour?"
• "Why is my S3 sink connector failing? Show the error logs."
• "Get fix recommendations for the postgres-source connector."

Get started

Ready to try the Confluent Cloud MCP server?

• Try it on watsonx Orchestrate Catalog → watsonx Orchestrate
• Don't have an instance yet? → Start a free WxO Trial
• Learn more about IBM Confluent → Confluent