Quick Start

From zero to a fully-loaded project memory in five minutes. Make sure you have completed the installation first.

1

Configure the MCP server

Add Engrams to your MCP client config (see Installation) and restart your IDE.

2

Scaffold the strategy file

Run from your project root:

engrams init --tool roo   # or cline, cursor, windsurf, claude-code…

This creates the custom-instructions file that tells your AI how to use Engrams.

3

Bootstrap from existing docs (optional)

If you already have a README.md, architecture doc, or similar, ask your AI to import it into Engrams: 

You: Initialize the Engrams database by reading README.md
     and docs/ARCHITECTURE.md

AI:  [ENGRAMS_ACTIVE] Engrams initialized.
     Architectural patterns stored.
     Coding standard decisions recorded.

     What would you like to work on?
4

Log your first decisions

Start a conversation and tell your AI about important choices:

You: The team has decided to use PostgreSQL as our primary database
     because we need ACID guarantees and complex query support.

AI:  Decision logged to Engrams:
     Decision #1: Use PostgreSQL for primary database
     Rationale: ACID guarantees, complex queries, mature ecosystem
     Tags: database, architecture
5

Watch context surface automatically

Now ask your AI to implement something:

You: Add a new /users endpoint to the API

AI:  I'll help with that. Retrieving project context...

     Found Decision #1: PostgreSQL is the primary database
     Found Decision #7: JWT auth required on all routes
     Found Pattern #2: Structured error handling

     Implementing with SQLAlchemy, JWT middleware,
     and proper error responses...

The AI found those decisions automatically — you didn't have to mention them.

What to try next

🔗 Codebase Bindings Bind decisions to file paths so context follows the code ⚖️ Team Governance Enforce team-level architectural mandates 📊 Knowledge Dashboard Visually explore your project's knowledge graph 💻 CLI Reference All engrams commands