Confluent Cloud managed MCP server

Name: Confluent Cloud managed MCP server
Author: confluent.cloud

Confluent-hosted MCP endpoints for Kafka clusters, connectors, schemas and Flink SQL in Confluent Cloud

Official · confluent.cloud Remote (hosted) · Streamable HTTP Cloud platforms Databases DevOps and monitoring

Website

40 popularity score

Connect your client

Remote (hosted)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Its sign-in lets no client register itself: a client signs in only with an app (a client ID) registered with its maker first, so these snippets work only after that step.

Claude Codeterminal

claude mcp add --transport http \
confluent-mcp-global \
https://api.confluent.cloud/mcp/v1 \
--header "Authorization: Basic <ENCODED_CREDENTIALS>"

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "confluent-cloud-managed": {
      "url": "https://api.confluent.cloud/mcp/v1"
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "confluent-cloud-managed": {
      "type": "http",
      "url": "https://api.confluent.cloud/mcp/v1"
    }
  }
}

What it does

Confluent runs this MCP server itself as part of Confluent Cloud; there is nothing to install. An assistant connects over streamable HTTP and can look through environments and Kafka clusters, check connector health and logs, read topic settings, sample messages and Schema Registry subjects, query metrics, and validate, run and debug Flink SQL statements.

It is really two endpoints. The global one at https://api.confluent.cloud/mcp/v1 serves 17 tools for environments, clusters, connectors and metrics. A regional one, whose address carries the cloud provider, region and organization ID, serves 15 tools for topics, schemas and Flink SQL. Confluent suggests adding both to a client; the global one alone is enough for discovering environments and clusters.

Most tools only read. Four change things: restart_connector, update_connector_config, create_flink_statement and delete_flink_statement. The server asks the client to confirm before those run, but the client decides whether that prompt is shown. Two connector tools return summaries and fix suggestions written by AI models, which Confluent says to review before acting on. Every tool call is recorded in the Confluent Cloud audit log under a method name starting with mcp.tools/call., attributed to the key that made it.

Confluent also publishes a separate open-source server, mcp-confluent, that you run yourself from npm.

What it needs

A Confluent Cloud account and an API key for one organization, sent as HTTP Basic auth (Base64 of key:secret in an Authorization header). A global API key works on both endpoints; a Cloud API key works only on the global endpoint; a Flink API key only on the regional one. Resource-scoped keys (a cluster or Schema Registry key) are refused. A client that speaks streamable HTTP. Private clusters need a private regional endpoint, a support ticket to enable it, and PrivateLink Gateway.

Limits

It cannot create or delete connectors, and cannot create, change or delete clusters or topics. consume_kafka_messages reads at most ten messages; get_connector_logs looks back 72 hours at most; get_flink_statement_exceptions returns the 10 latest. The regional endpoint exists only where Confluent Cloud Flink runs. A public regional endpoint cannot reach private clusters. Rate limits are those of the Confluent Cloud API. Each key covers one organization, so several organizations need several connections. update_connector_config cannot remove a field.

Who it suits

Teams already on Confluent Cloud who want an assistant to explain a failing connector, check lag or throughput, peek at topic data, or debug Flink SQL without running anything locally. Since it is hosted by Confluent and mostly read-only, it fits well for investigation. Anyone on self-managed Kafka or Confluent Platform, or who needs to create topics, clusters or connectors, should look elsewhere; Confluent's own self-run server, mcp-confluent, is a separate option for running it on your own machine.

Tools 32

list_environments: [global] List the Confluent Cloud environments the key can see, with IDs and governance package
read_environment: [global] Show the details of one environment
list_clusters: [global] List Kafka clusters in an environment with type, cloud, region and availability
read_cluster: [global] Show one cluster's details and endpoints
list_connectors: [global] List a cluster's connectors with name, type and status
get_connector_config: [global] Return a connector's full configuration and task list
get_connector_status: [global] Return a connector's runtime and per-task status, with error traces
get_connector_logs: [global] Fetch recent connector logs; ERROR level by default, WARN and INFO optional, 72 hours back at most
get_connector_offsets: [global] Show partitions and offsets a connector has consumed
get_connector_metrics: [global] Fetch time-series metrics for a connector
get_connector_error_summary: [global] Return a model-written summary of a connector's errors
get_connector_fix_recommendations: [global] Return model-written fix suggestions for a failing connector
restart_connector: [global] Restart a connector and its tasks (changes resources)
update_connector_config: [global] Change named fields of a connector's configuration, leaving the rest as is (changes resources)
list_metrics: [global] List metric descriptors for Kafka, connectors, compute pools, Flink statements, ksqlDB and Schema Registry
query_metrics: [global] Query a cluster's time-series metrics such as throughput or consumer lag, grouped by topic or group
get_flink_statement_metrics: [global] Fetch time-series metrics for a Flink SQL statement: throughput, backpressure, lag, state size
list_kafka_topics: [regional] List the topics in a cluster
describe_kafka_topic: [regional] Show a topic's configuration, partition count and replication factor
consume_kafka_messages: [regional] Read one to ten sample messages from a topic, optionally at a given partition and offset
list_schema_subjects: [regional] List Schema Registry subjects with version, schema ID and type
read_schema_subject: [regional] Return a subject's full schema, version and references
describe_flink_table_schema: [regional] Show a Flink table's columns, types, keys, watermarks and connector options
validate_flink_sql: [regional] Compile a Flink SQL statement without running it and return any error
create_flink_statement: [regional] Create and start a Flink SQL statement in a compute pool (uses compute, may write data)
list_flink_statements: [regional] List Flink SQL statements in an environment with a short SQL preview
read_flink_statement: [regional] Show a statement's phase, scaling, warnings and timestamps; long SQL is cut
delete_flink_statement: [regional] Stop and delete a Flink SQL statement for good
check_flink_statement_health: [regional] Give one health verdict for a statement with the evidence behind it
detect_flink_statement_issues: [regional] Look for common statement faults such as out-of-memory, timeouts or state limits, with hints
get_flink_statement_exceptions: [regional] Return a statement's 10 most recent exceptions, with long traces cut
get_flink_statement_profile: [regional] Show a running statement's task graph and flag the slowest task

Popularity

40 popularity score, on our list (the bar is 36)

GitHub stars	–	0 points
Official server	Yes	8 points
Hosted by its vendor	Yes	32 points

Measured 2026-11-07. How the score works.

Where it is published

Hosted	`https://api.confluent.cloud/mcp/v1` Streamable HTTP · OAuth sign-in

What it can reach

Runs	On the maker's servers (remote)
Reaches	Cloud resources, Your account at a service, A database
Changes things	Yes: it can write, send, create or delete
Sign-in	OAuth sign-in
Licence	The maker's terms (a hosted service)

The code runs on Confluent's servers, not your machine. It signs in with a Confluent Cloud API key and secret sent as HTTP Basic auth; what it can see and change follows that key's role-based permissions, the same as the Confluent Cloud API and CLI. Four tools change resources: restart_connector, update_connector_config, create_flink_statement and delete_flink_statement (deletion cannot be undone). The server asks the client to confirm these, but a client set to auto-approve may approve without showing you. consume_kafka_messages reads real topic data. Calls are logged in the Confluent Cloud audit log. A private regional endpoint does not enforce network isolation between private clusters inside one environment; only permissions stop such requests.

Over time

Popularity score 40 +0

4040

15 Oct 20267 Nov 2026

What is rising across the directory

Sources

Confluent's docs for the managed MCP servers: endpoints, tools, authentication, client setup, limitations, private networking

Page checked 2026-10-15.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![Confluent Cloud managed MCP server on EveryMCPServer](https://everymcpserver.com/badge/confluent-cloud-managed-mcp-server.svg)](https://everymcpserver.com/servers/confluent-cloud-managed-mcp-server)

Connect your client

Add in one click

Or add it by command or config

What it does

What it needs

Limits

Who it suits

Tools 32

Similar servers

Confluent MCP Server

Azure MCP Server

StackQL MCP Server

Render MCP Server

MCP DigitalOcean Integration

Heroku Platform MCP Server