ContextCake and data catalogs

ContextCake serves written project context to AI tools: decisions, instructions, runbooks, and notes, with sources and disagreements attached. Data catalogs focus on discovering and understanding data assets. These can be complementary parts of an engineering workflow.

What ContextCake does today

• Reads configured Markdown folders, OKF bundles, GitHub sources, and foreign MCP sources that implement ContextCake’s graph contract.
• Resolves matching sections across layers and preserves dissenting values.
• Shows source health and coverage; measures retrieval in native diagnostics.
• Applies recorded source policies only while their evidence and scope remain valid.

Source precedence is a configured selection rule. It is not a confidence score, a claim of factual correctness, or an authorization policy. The permissions of the underlying folders, repositories, and processes still matter.

Authoritative files and local copies

Source files remain authoritative. ContextCake maintains derived indexes, may cache source responses on disk, and can clone repositories for desktop sources. Its file editor and explicit resolution actions can change configured source files; recorded source policies preserve the original documents.

This is different from promising that no content is copied. Removing a source from the active cascade is also different from revoking access to a repository. Read the trust boundary.

Integration requires an adapter

A system having an MCP server does not make it a compatible ContextCake source. The current foreign-source adapter requires list_nodes and get_node with the specified shapes. Connecting a catalog such as OpenMetadata would require a verified translation into that contract. No production catalog connector ships with ContextCake today. See foreign MCP sources.

A small dbt artifact import is under assessment. It would bring model descriptions and dependencies into a generated layer while keeping human annotations separate. There is no dbt importer, warehouse connection, column-level lineage service, or new graph database in the current increment.

For catalog capabilities and deployment requirements, consult the catalog’s own current documentation. ContextCake does not replace data governance, warehouse permissions, data quality execution, or an organization’s catalog.