In 1860s America, trains couldn’t cross state lines.
Ohio ran a 4’10” gauge. The South ran 5′. Every freight car that hit a border had to be unloaded by hand and reloaded onto a different train. Then, over two days in 1886, more than 11,000 miles of track were relaid to a single standard gauge in a coordinated push. A car in Atlanta could roll straight through to Chicago. The network just worked.
Data tooling is still stuck at that border.
You’ve got semantic views in Looker, LookML in your repo, dbt semantic models, Snowflake Semantic Views, DAX in Power BI, and a legion of shadow spreadsheets that somehow became the source of truth for three teams. Each one runs its own gauge. Each analysis can only ride one track at a time.
Today, Zenlytic is shipping MCP Connectors.
The fear every data leader already has
Every data leader has the same reaction when they hear about a new AI agent: “Great. Now I have to maintain yet another context layer.“
That fear is valid. Your business logic is scattered across Looker, dbt, Snowflake, and Power BI, each holding definitions someone fought hard to get right. Drop in an agent that starts from zero, and those definitions drift. You end up with wrong answers, manual synchronization, and in the worst case, the wrong email is sent to a real customer.
The definitions already sitting in your tools are a goldmine. New agents should start there, not from zero.
How it works
Zoë connects directly to your existing tools (usually as an MCP client), reads the context already in place, and sets herself up. No rebuild. No migration. No second copy to maintain.
1. Ingestion. Zoë pulls in definitions, relationships, and lineage from your source systems on a schedule. The sync stays current on its own.
2. Use context from anywhere. Maybe your conversion metrics are in LookML, but your sales pipeline is in dbt. Zenlytic will retrieve (and even combine) context from both. Verified fields stay separate from dynamic ones, so you always know what is governed versus what the agent generated on the fly.
3. Context that learns. As users ask real questions, Zoë fills in gaps and refines her own context to standardize it over time. The data team has full visibility into what the business is actually asking. Learn more about our context layer.
4. Writeback. Verified context flows back optionally flows back to the source, either written directly or proposed as a pull request. The layers you already maintain stay in sync with what Zoë has learned.
Can’t we all just get along?
Zenlytic doesn’t want to be a single source of truth. We know it’s not even possible at enterprise scale. We just want your users to get the answers they need, and for data teams to connect the dots wherever they live.
You don’t have to turn everything on at once. Start with ingestion so Zoë is useful on day one. Add writeback once you trust the loop.
Every question makes the next one smarter.