Research & KnowledgeEvent: August 31, 2026
Cassis tested the method on a revenue domain from GitLab’s public analytics project, assembling context for 2,573 of 2,641 columns and surfacing 45 unresolved questions.
Published
Signal category
Research & Knowledge
Quote
“We tested the method on a revenue domain from GitLab’s public analytics project. It assembled context for 2,573 of 2,641 columns”
— Matthieu Blandineau|Cassis team
Company
Cassis
Context maintenance for analytics agents. Current, tested, approved.
- Industry
- Data Infrastructure and Analytics
- Company size
- 5 employees
Cassis keeps the context analytics agents rely on current as schemas and definitions change. Context lives in a Git repository you own, is served to agents over MCP, and changes through tested, reviewed pull requests. Questions and source changes surface gaps and drift. Your data team remains the approval gate. See the loop at https://getcassis.com/product/.