How Tessel works
Four steps, all inside the tools your team already uses. Tessel reads metadata and code; your team stays in charge of every change.
From failure to fix
-
Watch
Tessel reads run logs and metadata from your orchestrator and warehouse. Read-only, never query results.
-
Trace
It walks the lineage graph to the first step that went wrong and groups everything downstream into one incident.
-
Propose
If it knows the fix, it writes the change and a test, then opens a pull request on GitHub.
-
Review
Your team reviews and merges as usual. If Tessel isn’t sure, it explains the cause and leaves the fix to you.
Built on three decisions
These choices shape everything Tessel does, and they’re why security teams say yes.
-
Metadata, never data
Tessel reads run logs, compiled code and warehouse metadata such as row counts and load times. Query results never leave your warehouse.
-
Lineage first
Every alert, fix and freshness check starts from one graph of your sources, models and columns. That is how Tessel tells a cause from a symptom.
-
Your review, your merge
Fixes arrive as pull requests with a reason and a test. Branch protection and CI apply as usual, and Tessel never merges.
What it fixes, and what it explains
Tessel proposes a fix only when the intent is clear from the code and its history.
-
Renamed or retyped columns
Updates the select, adds a cast if needed and a test on the new column.
-
New values in a category
Adds the value to accepted-values tests and mappings, and flags it in the description.
-
Broken refs after a refactor
Points the ref at the renamed model and checks nothing else still uses the old name.
-
Late rows in incremental models
Widens the lookback window and adds a test for late-arriving data.
-
Missing grants and permissions
Names the missing privilege and proposes the change to your grants file.
-
Business logic questions
When the right answer needs a decision, such as how to count a new product line, Tessel explains the cause and stops.
Questions about how it works
Does Tessel need write access to our warehouse?
No. It reads metadata and run logs. Fixes arrive as pull requests that your team reviews and merges.
Where does our data go?
Nowhere. Query results never leave your warehouse. Tessel stores run metadata in its UAE or Saudi region, your choice.
How long does setup take?
Most teams connect their orchestrator and repository in under half an hour.
What happens when Tessel gets it wrong?
You close the pull request, optionally with a reason. Tessel learns from it, and you can switch off any fix pattern for a project.
Try it on a real failure
Start free with up to 25 models, or let us show you on a copy of your project.