Every source, check and agent task —
in one place.
Convergate is the operations hub behind our own products. It keeps one catalogue of data sources, shows where every dataset came from, runs and supervises recurring checks, and makes AI-agent work traceable — so a team always knows what needs attention and why.
Many streams, one gate.
Source catalogue
One register of projects and data sources, with owners, rights and freshness — instead of knowledge spread over repositories and chats.
Data lineage
Follow every dataset and model run back to its inputs, code version and execution, based on the OpenLineage standard.
Supervised checks
Recurring checks and monitors run as separate, bounded processes with time, memory and CPU limits. A failing check stays visible until it is fixed.
Attention first
A risk agenda answers one question: what needs my attention? Every alert explains what is going on, why it matters and the next step.
Accountable AI agents
Topics, handoffs and task records for Claude, Codex and other agents. Preparation, waiting and finished work stay distinct — agents never take over silently.
Evidence & guides
Per-release evidence, reproducible audit exports and administrator and user guides that are checked daily against the software version.
Metadata, not your data.
Convergate works with descriptions of your sources and runs, not copies of your product data. It runs locally or on your own infrastructure, and starts empty: no configuration, credentials or data from other installations.
- Runs standalone, on your own machines
- Fail-closed: a check that cannot start says so
- Secrets are filtered from worker environments
- No automatic approval — changes need an explicit review
- Every task leaves a compact, verifiable record
Running several products or data pipelines?
Convergate is in use internally and opening to a small number of early partners.