Agent-native ingest
Connect Claude, ChatGPT, or Cursor to the MotherDuck MCP Server. Describe the source and your agent writes the connector, deploys it, and sets the schedule. Guided by best practice, Python expertise optional.
FlightsPublic Preview
Build durable data pipelines in minutes
Python jobs built by your favorite agent and deployed in MotherDuck. Ingest, enrich, and transform data without expensive ETL.
Join the Flights launch livestream!
Join the team behind Flights for a getting-started walkthrough and live Q&A. Register now to save your seat.
From natural language to a running pipeline
Bring your sources and your agent. Flights handle the infrastructure: Python runtime, scheduling, secrets, logging, and versioning.
Python pipelines meet sub-second analytics
Anything you can pip install, you can run in a Flight. Ingest from databases, SaaS tools, or files. Then visualize in a Dive, all in a single agent session.
Built on a general-purpose Python runtime
Flights run Python on a dedicated runtime. Bring any pip-installable package, shell out to system binaries through subprocess, and run the same primitive across ingest, transforms, backfills, and scheduled maintenance.
Backed by SQL table functions
Create and manage Flights from any MotherDuck client using SQL table functions. Schedule runs, fetch logs, version your code, and trigger Flights on demand with table functions like md_create_flight, md_run_flight, and md_flight_runs.
From raw data to answers in a single thread
Flights ingests and transforms. Dives explores and visualizes. The same agent drives the whole loop through one MCP server, taking raw data to answers in a single chat thread.
Get Started with Flight Plans
FAQS
Flights are scheduled Python jobs that run on a dedicated Python runtime inside MotherDuck. MotherDuck handles the runtime, the cron schedule, secrets, versioning, and run history. The most common use is data ingestion, replacing ELT tools or hand-rolled Python scripts, and you can bring any pip-installable package.
Three paths, all backed by the same primitive:
- Agent + MCP — connect Claude, ChatGPT, Cursor, or any MCP-aware agent to the MotherDuck MCP server, describe the source, and the agent writes, deploys, and schedules the Flight.
- MotherDuck UI — open the Flights tab, paste in Python, set a cron, and save.
- SQL table functions — call
md_create_flight(...)from a SQL query to create and manage Flights programmatically.
No. Agents are the fastest path from "ingest this source" to a deployed pipeline, but you can write Flights yourself in the MotherDuck UI or create and manage them through SQL table functions like
md_create_flight and md_run_flight. Pick whichever workflow fits your team.The MotherDuck MCP server exposes tools that let any MCP-aware agent inspect your workspace, query your data, and create or update Flights on your behalf. When you ask an agent to ingest a new source, it uses the MCP server to understand your existing schema, write a Python pipeline against your data, deploy the Flight, and set the schedule. The same MCP surface drives Dives, so a single agent thread can ingest, transform, and visualize end-to-end.
Flights are billed for the compute consumed by the Python runner, by the second, plus standard Duckling rates during the MotherDuck load phase. Flights are included on Business and Enterprise plans. See the pricing page for current rates.
Take flight with your data
Your agent writes the pipeline. MotherDuck schedules, runs, and monitors it.
