Docs / Introduction Learn
Comparisons
Where EKOS fits next to RAG, vector databases, knowledge graphs, MCP and data catalogs.
EKOS is not competing with any one of these; it sits across several.
| Technology | Primary purpose | How EKOS relates |
|---|---|---|
| RAG | Retrieve relevant content for a prompt | EKOS can answer from typed evidence sets instead of raw chunks; ekos ask is a grounded, citation-validated pipeline |
| Vector DB | Store and retrieve embeddings | EKOS has an optional brute-force cosine vector index fused with BM25 via reciprocal-rank fusion |
| Knowledge graph | Model relationships | The CKM is a graph, but built by deterministic compilation with per-edge evidence, not hand-curated |
| MCP | Connect agents to tools and context | EKOS serves MCP; it is one of the things an MCP client can connect to |
| Data catalog | Describe data assets | EKOS recovers table/column/pipeline metadata from code and DDL, and links it to the code and documents that use it |
| EKOS | Compile heterogeneous engineering knowledge into an agent-queryable knowledge layer | — |
When EKOS is the wrong tool
- You need to search the contents of documents at scale with no interest in structure — a search engine is simpler.
- You need live operational data. EKOS describes systems; it does not query them (the one gated exception is the opt-in ClickHouse NL-to-SQL tool).
- You need a hand-authored ontology with formal reasoning. EKOS's model is recovered, not authored.