Start a Kafka Connect worker with a worker.properties file specifying bootstrap.servers, plugin.path, and key/value converters
Install connector JARs by placing them under the plugin.path directory and restarting the worker
Register a source connector via POST to /connectors with JSON body containing name, config.connector.class, tasks.max, and connector-specific settings
Register a sink connector similarly, specifying topics and the target system connection parameters
Monitor connector status with GET /connectors/YOUR_CONNECTOR_NAME/status and tasks with GET /connectors/YOUR_CONNECTOR_NAME/tasks/0/status
Known gotchas
Converter mismatches between source and sink (e.g., one using Avro, the other JSON) silently corrupt records; always specify key.converter and value.converter explicitly per connector
tasks.max is a hint, not a guarantee; the actual number of tasks is bounded by the number of source partitions or tables the connector discovers
Restarting a connector does not restart its tasks automatically in all versions; POST to /connectors/YOUR_CONNECTOR_NAME/tasks/0/restart separately if a task is in FAILED state
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp