Now with Dagster support
Data pipelines that fix themselves
Tessel watches every run, finds the step that broke and opens a pull request with the fix, usually before anyone notices.
Running in production at
- Mizan Pay
- Wasla
- Nasma Cooling
- Sahel Bank
- Rawda Health
- Qanat
- Manzil Homes
From red to green without a war room
Tessel connects to the tools you already run and works in your existing review flow.
-
Finds the broken step
Lineage-aware alerts point at the model that failed, not the forty that depend on it.
-
Proposes the fix
A pull request with the change, the reason, and a test that proves it.
-
Learns your stack
dbt, Airflow, Dagster, Snowflake, BigQuery and Postgres, with Slack and GitHub.
-
Keeps dashboards fresh
Freshness checks on every table your team actually looks at.
- of failures fixed without a human
- 83%
- median time to a proposed fix
- 11 min
- pipelines monitored
- 2,400
What changes for data teams
All customer stories-
WaslaSame-day delivery, 420 dbt models
Wasla turned Monday morning firefights into a pull request to approve
Weekend failures used to cost the data team half of every Monday. Now most arrive fixed.
- Monday incident time
- 4 h → 20 min
- of failures fixed by pull request
- 91%
-
Nasma CoolingDistrict cooling, BigQuery
Nasma Cooling found three weeks of silently dropped rows on day one
An upstream schema change had been dropping meter readings without a single failed run.
- silent row drops since launch
- 0
- schema changes caught in the first quarter
- 37
-
Rawda HealthClinic network, regulated data
Rawda Health passed security review in a week and caught a real failure the same afternoon
A healthcare data team needed monitoring that never touches patient data.
- to connect
- 20 min
- fewer alerts to triage
- 70%
Teams that sleep through the night
-
The pull requests read like a colleague wrote them: what broke, why, and a test that proves the fix. We merge about four in five as they are.
Omar Al SuwaidiStaff Data Engineer, Mizan Pay
-
Our alert volume dropped by two thirds in the first week, because forty downstream failures now arrive as one incident.
Sneha PillaiData Platform Manager, Qanat
-
Freshness checks learned our load schedule on their own. Finance stopped messaging me at 9:05 every morning.
Hannah ClarkeAnalytics Lead, Manzil Homes
From the blog
All posts-
Why one failure sends you forty alerts
Alerts follow the shape of your DAG, not the shape of the problem. Here is how we group them by cause.
-
What 10,000 automatic fix pull requests taught us
Which failures a machine should fix, which it shouldn’t, and why the test matters more than the patch.
-
Schema drift: the silent failure your tests miss
When a source changes shape, pipelines often keep running and quietly produce the wrong numbers.
See it on your pipelines
A 30-minute demo on a copy of your own project. No slides. Or start free with up to 25 models.