Engrams is a governance-aware, context-intelligent memory platform built on MCP. Give your AI agents structured, searchable, and automatically surfaced project knowledge — without burning every token on context.
{
"mcpServers": {
"engrams": {
"command": "uvx",
"args": ["--from", "engrams-mcp", "engrams-mcp",
"--mode", "stdio"]
}
}
}Store decisions, patterns, progress, and custom data in a queryable SQLite knowledge graph.
Vector embeddings enable meaning-aware retrieval — not just keyword matching.
Two-layer scope hierarchy prevents individual choices from contradicting team-level architecture.
Bind decisions to file paths. When AI edits a file, the right context surfaces automatically.
Token-aware scoring loads only the most relevant items — keeping costs low as your knowledge base grows.
Browser-based explorer with interactive D3 graph visualization and optional local Ollama chat.