The bet made sense on paper. Simple interfaces are easier for someone starting out but they trade away something bigger: room to grow. The moment you wanted to do something the UI hadn’t been built for, the difficulty didn’t creep up, it jumped. The code was never the hard part, the environment around the code was.
AI changes that trade: it’s good at turning an idea into working code. Give it a safe, constrained place to run that code, and the old choice between “simple for beginners” or “powerful for experts” stops being a choice at all. Someone new can just ask for what they want. An engineer can still drop straight into the code and nobody has to pick a lane.
That’s the shift behind this release. The work has moved into the AI coding agent already sitting inside your own repository. Asking you to leave it, log into a separate console, and go read documentation is friction you can measure in stalled trials and abandoned rollouts. Meltano stopped asking.
We listened to our users, and here’s what they kept telling us:
“As an analytics engineer working in my AI editor inside my own repository, I want the platform’s knowledge available as agent context, so I can go from an empty repo to a working, ingesting pipeline without ever opening the UI, reading external docs, or leaving the environment I’m comfortable in.”
Meltano’s Knowledge, Seeded Into Your Repo, Not Locked Behind a Login
When a team provisions a Meltano workspace, the platform already creates a git repository for it. Here’s what’s new: from day one, that repository comes seeded with a curated knowledge kit that turns any AI agent you open in it into a Meltano expert. Zero setup. No editor lock-in. It’s our own product knowledge, packaged so anyone can use it as a plugin, in whichever editor you already work in.
How to Go From Empty Repo to Working Pipeline
1. Open your workspace repo in your agent. The knowledge kit is already there. Nothing to install or configure first.
2. Ask for the outcome you want. Something like “load this CSV into Postgres” is enough. The agent reads the knowledge base, works out you need `tap-csv` extracting into `target-postgres`, and drafts the pipeline configuration itself.
3. Let it flag what’s secret, without seeing it. The agent scaffolds a placeholder for every credential the pipeline needs, things like a database password, but never touches or stores the real values.
4. Add your real credentials directly on the platform. One short, deliberate step per secret, kept off your machine and out of your repo.
5. Push to ship. Your repo is the source of truth, so the platform reconciles the configuration and your pipeline goes live on hosted infrastructure. No separate promotion step in a UI.
6. If something breaks, ask the agent, not a log viewer. It reads the failure, explains it, and proposes a fix without you leaving your editor. This runs alongside Meltano’s existing platform diagnostics, not instead of them.
Every fix worth remembering gets written back into your own repo. The workspace gets a little smarter every time someone touches it.
Except for One Thing: Your Secrets
This isn’t zero interaction with Meltano’s platform, and that’s deliberate. An API token or a warehouse password should never sit in an AI agent’s context or get committed to a repository. So the agent stops at that boundary, and you supply the real values directly and securely on the platform. One brief step per credential, in the one place a little friction is worth it.
What This Actually Saves You
Reading documentation and switching into a separate UI are the two things most likely to stall you before you’ve built anything. Put the knowledge beside the code instead, and both problems disappear. You don’t lose an afternoon working out which connector to use or where a setting lives. You go from an empty repo to a pipeline that’s actually running, in about the time it takes to describe what you want.
If your team is already building with an AI coding agent, the fastest way to see this is to try it on your own pipeline.