Conduit MCP. Check, debug, and manage your data pipelines through chat.

This isn't some fancy dashboard you gotta log into; it’s your AI client talking straight to the Conduit data pipelines. You manage complex data moves between systems just by chatting with your agent. It lets you inspect everything—health checks, connector lists, run logs—using plain text commands.

When you need to know what's running or what could run, you start here. First, you use list_workflows to pull a list of every data integration workflow ID that exists in Conduit. This list gives you the specific IDs you’ll need for everything else—you gotta reference those later.

To get the full scope of any given pipeline, you can call get_workflow. This returns detailed information about an entire workflow, showing you exactly what sources and destinations it's set up to handle. You can also check out what’s connected by running list_connections, which pulls a definitive list of every active source-to-destination link configured on your account.

If you need to know how that pipeline starts or ends, the system tracks it for you. Use list_available_sources and you get a comprehensive list of all supported databases and endpoints that can serve as the starting point for any data flow. Similarly, if you run list_available_destinations, you'll see every type of external system or service that can receive data from a pipeline.

When it comes to checking health, your agent handles it in real time. To check the operational status of a specific workflow at this moment, you use get_run_status. This returns the current status, timing details, and any error messages for a single, defined run. If you need to audit what went down over time, call list_workflow_runs to get the full execution history for that workflow, providing both statuses and timestamps for every past attempt.

If things are acting up or you gotta re-process some data, you don't wait around. You just tell your agent to run it using trigger_workflow. This manually kicks off a run for a specific data workflow ID when you need to test something out or force a reprocessing job.

It’s all about control. Need to map out the whole infrastructure? Use the tools to check every active connection, listing them with list_connections, and cross-reference that against the available components provided by list_available_sources and list_available_destinations. You'll get a complete picture of your entire data landscape. For instance, if you want to see what specific databases are ready for use, run list_available_sources; if you need to know which systems can accept the output, check list_available_destinations.

This lets you build out complex, reliable pipelines entirely through chat.

The whole process is built around transparency. You start by finding all workflows with list_workflows, then you use that ID to get deep details about it via get_workflow. If the pipeline is running smoothly, great; if not, immediately check its health using get_run_status. If there’s been an outage or a hiccup, don't sweat it.

You just ask for the history with list_workflow_runs to see exactly when things went sideways and why.

If you figure out the issue and need to test a fix, you can manually restart the job using trigger_workflow. The system tracks everything—the status, the timing, and the error codes—so you never have to guess what’s happening under the hood. You're not stuck clicking buttons in some clunky web dashboard; your agent handles all that heavy lifting for you, keeping you right in the chat window.