Quick Start
From zero to a fully-loaded project memory in five minutes. Make sure you have completed the installation first.
Agent Integration
Engrams provides a built-in instructions command designed to output the standardized instruction manual that teaches LLM agents how to interact with the Engrams CLI:
engrams instructions
Run this command in your project root and copy-paste the output directly into your project's AGENTS.md file, .clinerules, .cursorrules, or system prompt configuration. This teaches the AI agent exactly when and how to call the CLI to query context, log decisions, and update progress.
Interactive Walkthrough
Orient the agent on startup
At the start of a session, your AI agent runs engrams prime to fetch the project overview, active focus, decisions, and patterns all in one call:
You: Let's start working on the project.
AI: Sure, let me orient myself with the project context first.
[Running: engrams prime --budget 1000]
Loaded product context, active context, and recent decisions.
I see we are building a Rust CLI using SQLite, and the active task is implementing search. I'm ready!Bootstrap from existing docs (optional)
If you already have a README.md, architecture doc, or similar, ask your AI
to parse it and log the key facts to the Engrams database. Behind the scenes, the AI will execute CLI commands:
You: Parse my README.md and store our core architecture choices in Engrams.
AI: I'll extract the core architecture choices.
[Running: engrams decision log --summary "Use Node.js with TypeScript" --rationale "Team proficiency" ...]
Logged 3 architectural decisions successfully!Log a decision manually
When you and your AI agree on a design choice, have the AI log it to keep the database fresh:
You: We've decided to use SQLite with FTS5 for local search.
AI: Understood. Logging that decision.
[Running: engrams decision log --summary "Use SQLite with FTS5 for local search" --rationale "Fast, local, simple" --tags "search,db"]
Decision logged as ID 4.Retrieve context on demand
The AI agent checks Engrams at the start of a task or searches for topics to ensure it follows your conventions:
You: Write a search module for the database.
AI: Let me search Engrams for search-related decisions first.
[Running: engrams decision search "search"]
Found:
- Decision #4: Use SQLite with FTS5 for local search (Fast, local, simple)
I'll implement the search module using SQLite FTS5 as decided.Because the AI queried the CLI directly, it aligned with your stack choice instantly—without you needing to re-state the rules.