StreamNative MCP Server
About
Integrate AI agents with StreamNative Cloud resources and Apache Kafka/Pulsar messaging systems.
Details
- Author
- streamnative
- Categories
- Cloud Service, Infrastructure, Other, Automation
Jump to
Setup
Install StreamNative MCP Server in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/streamnative/streamnative-mcp-server
Follow the installation instructions in the repository README, then restart your MCP client.
Integrate AI agents with StreamNative Cloud resources and Apache Kafka/Pulsar messaging systems.
A Model Context Protocol (MCP) server for integrating AI agents with StreamNative Cloud resources and Apache Kafka/Pulsar messaging systems.
StreamNative MCP Server provides a standard interface for LLMs (Large Language Models) and AI agents to interact with StreamNative Cloud services, Apache Kafka, and Apache Pulsar. This implementation follows theModel Context Protocolspecification, enabling AI applications to access messaging services through a standardized interface.
The server currently negotiates MCP protocol versions2025-11-25,2025-06-18,2025-03-26, and2024-11-05. The default preference is2025-11-25, while older clients remain supported through protocol negotiation.
- StreamNative Cloud Integration:
- Connect to StreamNative Cloud resources with authentication
- Switch to clusters available in your organization
- Describe the status of clusters resources
- Kafka Admin operations (topics, partitions, consumer groups)
- Schema Registry operations
- Kafka Connect operations ()
- Kafka Client operations (producers, consumers)
- Pulsar Admin operations (topics, namespaces, tenants, schemas, etc.)
- Pulsar Client operations (producers, consumers)
- Functions, Sources, and Sinks management
- Read-only MCP resources for context, catalog, and bounded admin summaries
- Connect to StreamNative Cloud with service account authentication
- Connect directly to external Apache Kafka clusters
- Connect directly to external Apache Pulsar clusters
: The Kafka Connect operations are only tested and verified on StreamNative Cloud.
The easiest way to install streamnative-mcp-server is using Homebrew:
# Add the tap repository brew tap streamnative/streamnative # Install streamnative-mcp-server brew install streamnative/streamnative/snmcp
StreamNative MCP Server releases the Docker Image tostreamnative/snmcp, and it can be used to run both stdio server and sse server via docker command.
# Pull image from Docker Hub docker pull streamnative/snmcp
Seecharts/snmcp/README.mdfor Helm installation via the StreamNative chart repository.
# Clone the repository git clone https://github.com/streamnative/streamnative-mcp-server.git cd streamnative-mcp-server go mod tidy go mod download # Build the binary make
If you want to access to your StreamNative Cloud, you will need to have following resources ready:
- Access toStreamNative Cloud.
- StreamNative Cloud Organization
- StreamNative Cloud instance and cluster
- Service Account with admin role
- Download the Service Account Key file
# Start MCP server with StreamNative Cloud authentication bin/snmcp stdio --organization my-org --key-file /path/to/key-file.json # Start MCP server with StreamNative Cloud authentication and pre-configured context # When --pulsar-instance and --pulsar-cluster are provided, context mutation tools are disabled bin/snmcp stdio --organization my-org --key-file /path/to/key-file.json --pulsar-instance my-instance --pulsar-cluster my-cluster # Start MCP server with external Kafka bin/snmcp stdio --use-external-kafka --kafka-bootstrap-servers localhost:9092 --kafka-auth-type SASL_SSL --kafka-auth-mechanism PLAIN --kafka-auth-user user --kafka-auth-pass pass --kafka-use-tls --kafka-schema-registry-url https://sr.local --kafka-schema-registry-auth-user user --kafka-schema-registry-auth-pass pass # Start MCP server with external Pulsar bin/snmcp stdio --use-external-pulsar --pulsar-web-service-url http://pulsar.example.com:8080 bin/snmcp stdio --use-external-pulsar --pulsar-web-service-url http://pulsar.example.com:8080 --pulsar-token "xxx" # Start MCP server with stdio by docker with StreamNative Cloud authentication docker run -i --rm -e SNMCP_ORGANIZATION=my-org -e SNMCP_KEY_FILE=/key.json -v /path/to/key-file.json:/key.json -p 9090:9090 streamnative/snmcp stdio
# Start MCP server with SSE and StreamNative Cloud authentication bin/snmcp sse --http-addr :9090 --http-path /mcp --organization my-org --key-file /path/to/key-file.json # Start MCP server with SSE and pre-configured StreamNative Cloud context # When --pulsar-instance and --pulsar-cluster are provided, context mutation tools are disabled bin/snmcp sse --http-addr :9090 --http-path /mcp --organization my-org --key-file /path/to/key-file.json --pulsar-instance my-instance --pulsar-cluster my-cluster # Start MCP server with SSE and external Kafka bin/snmcp sse --http-addr :9090 --http-path /mcp --use-external-kafka --kafka-bootstrap-servers localhost:9092 # Start MCP server with SSE and external Pulsar bin/snmcp sse --http-addr :9090 --http-path /mcp --use-external-pulsar --pulsar-web-service-url http://pulsar.example.com:8080 # Start MCP server with SSE by docker with StreamNative Cloud authentication docker run -i --rm -e SNMCP_ORGANIZATION=my-org -e SNMCP_KEY_FILE=/key.json -v /path/to/key-file.json:/key.json -p 9090:9090 streamnative/snmcp sse
When running the SSE server with external Pulsar, you can enablemulti-session modeto support per-user authentication. In this mode, each HTTP request must include anAuthorization: Bearer <token>header, and the server will create separate Pulsar sessions for each unique token.
# Start SSE server with multi-session Pulsar mode bin/snmcp sse --http-addr :9090 --http-path /mcp \ --use-external-pulsar \ --pulsar-web-service-url http://pulsar.example.com:8080 \ --multi-session-pulsar \ --session-cache-size 100 \ --session-ttl-minutes 30
- Per-user sessions: Each user's Pulsar token creates a separate session
- LRU caching: Sessions are cached with LRU eviction when the cache is full
- TTL-based cleanup: Idle sessions are automatically cleaned up after the configured TTL
- Strict authentication: Requests without a validAuthorizationheader receive HTTP 401 Unauthorized
- Client connects to SSE endpoint withAuthorization: Bearer <pulsar-jwt-token>header
- Server validates the token by attempting to create a Pulsar session
- If valid, the session is cached and reused for subsequent requests
- If invalid or missing, server returns HTTP 401 Unauthorized
Note:Multi-session mode is only available for external Pulsar mode (--use-external-pulsar) and only works with the SSE server, not stdio.
Usage: bin/snmcp [command] Available Commands: stdio Start stdio server sse Start sse server help Help about any command Flags: --audience string The audience identifier for the API server (default "https://api.streamnative.cloud") --client-id string The client ID to use for authorization grants (default "AJYEdHWi9EFekEaUXkPWA2MqQ3lq1NrI") --config-dir string If present, the config directory to use --enable-command-logging When enabled, the server will log all command requests and responses to the log file --features strings Features to enable, defaults to all -h, --help help for bin/snmcp --issuer string The OAuth 2.0 issuer endpoint (default "https://auth.streamnative.cloud/") --kafka-auth-mechanism string The auth mechanism to use for Kafka --kafka-auth-pass string The auth password to use for Kafka --kafka-auth-type string The auth type to use for Kafka --kafka-auth-user string The auth user to use for Kafka --kafka-bootstrap-servers string The bootstrap servers to use for Kafka --kafka-ca-file string The CA file to use for Kafka --kafka-client-cert-file string The client certificate file to use for Kafka --kafka-client-key-file string The client key file to use for Kafka --kafka-schema-registry-auth-pass string The auth password to use for the schema registry --kafka-schema-registry-auth-user string The auth user to use for the schema registry --kafka-schema-registry-bearer-token string The bearer token to use for the schema registry --kafka-schema-registry-url string The schema registry URL to use for Kafka --key-file string The key file to use for authentication to StreamNative Cloud --log-file string Path to log file --organization string The organization to use for the API server --proxy-location string The proxy location to use for the API server (default "https://proxy.streamnative.cloud") --pulsar-auth-params string The auth params to use for Pulsar --pulsar-auth-plugin string The auth plugin to use for Pulsar --pulsar-token string The token to use for Pulsar --pulsar-cluster string The default cluster to use for the API server --pulsar-instance string The default instance to use for the API server --pulsar-tls-allow-insecure-connection The TLS allow insecure connection to use for Pulsar --pulsar-tls-cert-file string The TLS cert file to use for Pulsar --pulsar-tls-enable-hostname-verification The TLS enable hostname verification to use for Pulsar (default true) --pulsar-tls-key-file string The TLS key file to use for Pulsar --pulsar-tls-trust-certs-file-path string The TLS trust certs file path to use for Pulsar --pulsar-web-service-url string The web service URL to use for Pulsar -r, --read-only Read-only mode --server string The server to connect to (default "https://api.streamnative.cloud") --use-external-kafka Use external Kafka --use-external-pulsar Use external Pulsar --http-addr string HTTP server address (default ":9090") --http-path string HTTP server path for SSE endpoint (default "/mcp") --multi-session-pulsar Enable per-user Pulsar sessions based on Authorization header tokens (only for external Pulsar mode) --session-cache-size int Maximum number of cached Pulsar sessions when multi-session is enabled (default 100) --session-ttl-minutes int Session TTL in minutes before eviction when multi-session is enabled (default 30) -v, --version version for bin/snmcp
The StreamNative MCP Server supports enabling or disabling specific groups of functionalities via the--featuresflag. This allows you to control which MCP tools are available to your AI tools. Enabling only the toolsets that you need can help the LLM with tool choice and reduce the context size.
The StreamNative MCP Server allows you to enable or disable specific groups of features using the--featuresflag. This helps you control which tools are available to your AI agents and can reduce context size for LLMs.
Claude connector compatibility: admin tools that previously mixed read and write operations behind oneoperationparameter are exposed as separate read/write MCP tools, for examplekafka_admin_topics_readandkafka_admin_topics_write. Read tools includeannotations.readOnlyHint=true; write or side-effectful tools includeannotations.destructiveHint=true. In--read-onlymode, write/destructive tools are not registered.
Pulsar admin feature gates also register read-only MCP resources for the matching admin surface. These resources usepulsar://...URIs, return JSON snapshots, and stay separate from write-capable tools; seepulsar_resources.mdfor the supported URI templates and safety boundaries.
Note:When using--pulsar-instanceand--pulsar-clusterflags together, context mutation tools (sncloud_context_use_cluster,sncloud_context_reset) are automatically disabled since the context is pre-configured.
You can combine these features as needed using the--featuresflag. For example, to enable only Pulsar client features:
# Enable only Pulsar client features bin/snmcp stdio --organization my-org --key-file /path/to/key-file.json --features pulsar-client
npm install -g @modelcontextprotocol/inspector
# Inspect a stdio server mcp-inspector stdio --command "bin/snmcp stdio --organization my-org --key-file /path/to/key-file.json" # Inspect an SSE server mcp-inspector sse --url "http://localhost:9090/mcp"
The inspector provides a web interface where you can:
- View available tools and their schemas
- Test tool invocations
- Monitor server responses
- Debug connection issues
This server can be used with any MCP-compatible client, such as:
- Claude Desktop
- Other AI assistants supporting the MCP protocol
- Custom applications built with MCP client libraries
⚠️ Reminder: Please ensure you have an active paid plan with your LLM provider to fully utilize the MCP server. Without it, you may encounter the error:message will exceed the length limit for this chat.
{ "mcpServers": { "mcp-streamnative": { "command": "${PATH_TO_SNMCP}/bin/snmcp", "args": [ "stdio", "--organization", "${STREAMNATIVE_CLOUD_ORGANIZATION_ID}", "--key-file", "${STREAMNATIVE_CLOUD_KEY_FILE}" ] } } }
Please remember to replace${PATH_TO_SNMCP}with the actual path to thesnmcpbinary and${STREAMNATIVE_CLOUD_ORGANIZATION_ID}and${STREAMNATIVE_CLOUD_KEY_FILE}with your StreamNative Cloud organization ID and key file path, respectively.
Optionally, you can use docker image to start the stdio server if you haveDockerinstalled.
{ "mcpServers": { "mcp-streamnative": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "SNMCP_ORGANIZATION", "-e", "SNMCP_KEY_FILE", "-v", "${STREAMNATIVE_CLOUD_KEY_FILE}:/key.json", "streamnative/snmcp", "stdio" ], "env": { "SNMCP_ORGANIZATION": "${STREAMNATIVE_CLOUD_ORGANIZATION_ID}", "SNMCP_KEY_FILE": "/key.json" } } } }
Then configure Claude Desktop to use the SSE server:
{ "mcpServers": { "mcp-streamnative-proxy": { "command": "mcp-proxy", "args": [ "http://localhost:9090/mcp/sse" ] } } }
Note: If mcp-proxy is not in your system PATH, you'll need to provide the full path to the executable. For example:
- On macOS:/Library/Frameworks/Python.framework/Versions/3.11/bin/mcp-proxy
- On Linux:/usr/local/bin/mcp-proxy
- On Windows:C:\Python311\Scripts\mcp-proxy.exe
Please remember to replacehttp://localhost:9090/mcp/ssewith the right URL.
The Model Context Protocol (MCP) is an open protocol that standardizes how applications provide context to LLMs. MCP helps build agents and complex workflows on top of LLMs by providing:
- A growing list of pre-built integrations that your LLM can directly plug into
- The flexibility to switch between LLM providers and vendors
- Best practices for securing your data within your infrastructure
For more information, visitmodelcontextprotocol.io.
This section describes how to release a new version ofsnmcp.
- Generate a tag for the new version (see Versioning, below):
- build Go binaries for supported platforms
- archive the binaries
- publish a release to the github repository (ref)
- Stable:vX.Y.Z
- Pre-release:vX.Y.Z-rc.W
- Snapshot:vX.Y.Z-SNAPSHOT-commit
Licensed under the Apache License Version 2.0:http://www.apache.org/licenses/LICENSE-2.0
A Model Context Protocol (MCP) server that provides tools for AI, allowing it to interact with the DataWorks Open API through a standardized interface. This implementation is based on the Aliyun Open API and enables AI agents to perform cloud resources operations seamlessly.
Search and run capsules, execute pipelines, and manage data assets on the Code Ocean platform.
Production-grade MCP server for Databricks: SQL Warehouses, Jobs API, multi-workspace support.
An MCP server for managing Ray clusters, jobs, and distributed computing workflows.
Zero-setup MCP gateway securely connecting AI to your tools with authentication and workflows
Realtime and Historical Exchange Rate Data, FX pairs, bid/ask, OHLC and Crypto prices
Manage research data and compute with Globus.
Get your website online in seconds — just ask your AI assistant to publish it and Just Publish hands you a live link to share. No code, no setup, no hosting to figure out.
Skyvia MCP Endpoint is a no-code gateway that makes your business data accessible to AI assistants in real time.
Connect your business data to any MCP client. AgenticBI agents find the right data, join across systems, run queries, build dashboards, generate reports, and automate analytics workflows. No SQL required.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
