S2 StreamStore MCP Server

An MCP server for S2 StreamStore built with Concierge progressive tool disclosure.

Instead of exposing all S2 API tools at once, the server guides AI agents through a natural workflow:

account → basin → stream

At each stage, only the relevant tools are visible.

Stages

Stage	Tools	Description
account	list_basins, create_basin, select_basin, list_access_tokens, issue_access_token, revoke_access_token, account_metrics	Manage basins and access tokens
basin	get_basin_config, reconfigure_basin, delete_basin, list_streams, create_stream, select_stream, basin_metrics, go_back_to_account	Manage streams within a basin
stream	get_stream_config, reconfigure_stream, delete_stream, append_records, read_records, check_tail, stream_metrics, go_back_to_basin	Read/write records and manage stream config

Setup

pip install -e .

Configuration

Set your S2 access token as an environment variable:

export S2_ACCESS_TOKEN="your-token-here"

Run

stdio (for Cursor, Claude Desktop, etc.):

python main.py

MCP Client Configuration

Add to your MCP client config (e.g. ~/.cursor/mcp.json):

{
  "mcpServers": {
    "s2": {
      "command": "python",
      "args": ["/path/to/mcp/main.py"],
      "env": {
        "S2_ACCESS_TOKEN": "your-token-here"
      }
    }
  }
}