Conduktor MCP Turn natural language into Kafka expertise.

Q: Can we disable MCP entirely?

Yes. Set `enable_mcp: false` and the endpoint doesn't exist.

Build with confidence. Faster incident resolution, governance insights across every dimension, and full visibility into your Kafka investment.

conduktor-mcp — cluster audit

> Cross-reference my cluster health, schema coverage, and replication risks. What should I fix this sprint? Cluster Health: 88% — 19 topics, 108 partitions, 5 schemas P0 — Fix now payment-events is the only topic with RF=1. If that broker goes down, you lose payment data. P1 — This sprint payroll-employee producing 40 msg/sec with zero consumers. ~3.4M messages/day going unread. ──────────────────────────────────────── P0 Fix: RF→3 on payment-events P1 Fix: Investigate payroll consumers

LLM Reasoning Meets Live Kafka Data

MCP combines AI reasoning with your live infrastructure and deep Kafka domain knowledge, unlocking capabilities that weren't possible before.

For Developers Understand how pipelines work, trace dependencies, and build new services with full context across clusters, schemas, and consumer groups.

For Platform Engineers Reason across every dimension of your Kafka infrastructure at once, without missing context or building ad hoc views.

For Engineering Leaders Know what your Kafka investment is hiding. Surface cost savings, uncover untapped capacity, and turn insights into opportunities.

Prompt Engineering for Kafka

MCP connects clusters, topics, schemas, consumer groups, and configurations in a single conversation, so you can understand how your applications and infrastructure interact and build with confidence, without losing a second.

Incident Investigation

"Orders-processor has been lagging for 2 hours. Check schema versions, topic configs, and partition assignments. What's the likely root cause?"

Configuration Audit

"Review my topic configurations against best practices for a 50k msg/sec workload. Flag anything suboptimal."

Compliance Evidence

"All topics containing PII, their retention, encryption, and consumer access. Format for our quarterly audit."

Onboarding and Discovery

"How does the order fulfillment pipeline work? Trace the data flow from orders-raw through all downstream consumers."

Migration Planning

"I need to migrate user-events to a new schema. What consumers depend on it? Safest migration path?"

Cost and Opportunity Discovery

"Which topics have retention policies that don't match their actual consumption patterns?"

MTTR from hours to minutesIncident investigation that would take an engineer hours of cross-referencing happens in a single conversation.

New engineers productive in daysOnboarding no longer depends on your senior engineers' availability. MCP explains your infrastructure in context.

Audit prep in minutes, not weeksCompliance evidence generated on demand, formatted for your auditors, from live data.

Every engineer, a Kafka expertDeep Kafka knowledge available to every team member, not just your most senior SREs.

How It Works

Architecture

Your Data Never Leaves Your Network

MCP is built into Conduktor Console. Connect Claude Code, Cursor, or any MCP-compatible client to your Console URL, and your Kafka data never leaves your infrastructure.

Built into Console: No separate deployment. MCP ships with your existing Conduktor Console instance
No data exfiltration: Kafka metadata is processed locally, never leaves your network
Token-based auth: Console Personal Access Tokens inherit user permissions
Authenticated on every request: Token validity checked on every call. Revoke a token and access stops immediately

MCP Architecture - AI assistant connects to Console within your network

Setup

Three Steps to Get Started

MCP uses Console's existing permissions model. No new access model to configure. A developer's AI assistant sees exactly what that developer already sees in Console.

Generate a token: Create a Personal Access Token in Console
Configure your AI assistant: Point Claude (or any MCP-compatible client) at your Console URL
Start asking questions: Natural language queries against your live Kafka infrastructure

Read the documentation →

Generate TokenConfigure ClaudeStart Querying

// claude_desktop_config.json
{
  "mcpServers": {
    "conduktor-console": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://console.acme-corp.com/api/mcp",
        "--header",
        "Authorization: Bearer ${CONDUKTOR_API_TOKEN}"
      ],
      "env": {
        "CONDUKTOR_API_TOKEN": "your-personal-access-token"
      }
    }
  }
}

Cost Discovery and Optimization

MCP explores your entire Kafka footprint to surface hidden costs no single dashboard can show, then synthesizes findings into prioritized optimization plans across every cluster, team, and dimension.

Discover Hidden Costs

Dashboards show predefined categories. MCP explores everything and follows the thread. "Walk through my entire Kafka footprint and flag anything misconfigured, underutilized, or disproportionately expensive. What am I missing?"

Compare Across Clusters

No UI compares clusters side by side. MCP reasons across your full infrastructure at once. "Compare my production and staging clusters. Are they proportionally sized, or am I overpaying for environments that don't need the same capacity?"

Quantify the Impact

Knowing the problem isn't enough. MCP projects what fixing it is worth. "If I right-size the top 20 over-provisioned topics and reduce retention where consumers read within minutes, what's the projected savings?"

Prioritize by Team

Turn findings into ownership. MCP attributes waste by team and ranks the highest-impact optimizations first. "Generate a cost optimization report by team: who owns the most waste, what should they fix first, and what's the estimated impact?"

Why Your Security Team Will Approve This

MCP doesn't create new access. It makes existing access more useful.
Same permissions, same network boundary.

Your Network, Your Data

MCP runs inside your Console instance. Kafka data never leaves your infrastructure to reach external services.

Console Permissions

Personal Access Tokens inherit Console RBAC. No new access model to configure. AI sees what the user already sees in Console.

Validated on Every Request

Token validity checked on every MCP call. Revoke a compromised token and access stops immediately.

Configurable Access

Disable MCP entirely with a single feature flag in Console. One config change and the endpoint doesn't exist.

Where does my data go?

The MCP server runs inside your Console instance, in your environment. Kafka metadata flows to the AI assistant on the user's machine. Nothing is sent to external AI training systems. Your data never leaves your network.

Who can use MCP?

Any Console user who can create a Personal Access Token. Tokens inherit the creator's Console permissions, so MCP access is scoped to exactly what that user can already see in Console.

Can we disable MCP entirely?

Yes. Set enable_mcp: false and the endpoint doesn't exist.

What if a token is compromised?

Revoke it in Console like any other token. The MCP server validates token validity on every request.

What's Next

MCP is a platform, not a point feature.
Today it surfaces intelligence. Tomorrow it acts on it, safely.

Topic Creation Assistance

AI-guided topic creation with best-practice defaults based on your workload.

Schema Evolution Workflows

Validate schema changes against consumers before deploying, with AI-powered compatibility analysis.

Automated Remediation

AI-recommended fixes with approval gates. Resolve common issues without context-switching.

See What Your Kafka Infrastructure Can Tell You

Book a demo and we'll show you MCP against a live environment. Incident investigation, compliance audits, cost discovery, and the questions you haven't thought to ask yet.

Book a Demo Read Documentation