engram integrates with the Codebase Context Specification — a standard for human + AI-readable project documentation stored in .context/index.md.
CCS defines a standard location (.context/index.md) for project documentation. Sections typically include Architecture, Decisions, Conventions, and Known Issues. Any AI tool can read this file to understand a project without digging through source code.
If your project already has a .context/index.md, import it into the knowledge graph:
engram init --from-ccsThis parses each section and maps bullet points to graph nodes:
| Section heading | Node kind |
|---|---|
| Architecture, Design, Conventions, Patterns | pattern |
| Decisions | decision |
| Issues, Problems, Known Issues | mistake |
| Everything else | concept |
All imported nodes get confidenceScore: 0.9 — human-authored context is treated as high-signal.
Export the knowledge graph as a CCS-format .context/index.md:
engram gen-ccsThis writes four sections:
- Architecture Patterns — pattern nodes with confidence >= 0.8, sorted by query frequency
- Decisions — all decision nodes, newest first
- Known Issues — mistake nodes sorted by how often they surface in queries
- Key Concepts — high-traffic concept nodes (queryCount > 0)
engram is the dynamic layer for CCS. Static .context/index.md files capture what you know at a point in time. engram makes that knowledge live — it grows as you code, surfaces the right nodes during AI queries, and tracks which decisions and patterns actually matter (via queryCount).
Static docs become a living knowledge graph.