LinkedIn·Wednesday, 26 August 2026·5h ago
CLI, API, or MCP? Three ways to connect data quality to your stack, and most teams pick by habit rather than by the job. The question that…
Maarten Masschelein
CEO & Co-Founder @ Soda | Data and Context Management for AI
CLI, API, or MCP?
Three ways to connect data quality to your stack, and most teams pick by habit rather than by the job.
The question that settles it: what triggers the work, and how often?
➨ 𝗖𝗟𝗜 — the pipeline triggers it.
Scheduled runs, CI checks, anything that repeats without a person in the room. Version-controlled and reviewable, in the same workflow as the rest of your code.
➨ 𝗔𝗣𝗜 — your own software triggers it.
You are building a product that needs quality checks inside it, and you want the results back as data you control.
➨ 𝗠𝗖𝗣 — a person or an agent triggers it, in plain language.
"Which of these datasets are ready to feed the model?" No syntax to remember. The agent reads your contracts and answers.
Most teams need two of the three. Almost nobody needs all three in month one.
Pick by what starts the job, not by what sounds most advanced.
🥸Hakim Elakhrass ran a full data quality lifecycle through Soda MCP (https://lnkd.in/eDPZRYS5): from an unmonitored dataset to a contract gating the pipeline, entirely no-code.
Recording in the comments.
♥ 10💬 2↻ 1
View on LinkedIn Cross-referenced
Related on the wire
You can do governance work in an afternoon that used to take months. Here are 3 things AI does for data quality that manual processes can't…
I am a data steward. How can I add AI into my workflow? Data stewards hold the business rules, but turning them into running checks…
Data analysts should not be the last line of defense for data quality. Yet that's where many teams end up. A data issue enters a pipeline…
Metadata is now growing beyond passive documentation. It has the potential to become "active". Active metadata is metadata that is…
MCPs have the potential to change data governance forever. Context is everything in data governance, and unfortunately for many teams its…
We already have data catalogs. Why do we need MCP? When an AI agent is about to act on a dataset, it shoud answer 4 questions first: → Is…