Project Onboarding & Agent Orientation

When a new developer joins your team or you start a fresh conversation with an AI agent, they need a quick way to get oriented. Engrams provides a structured onboarding flow using product context, active sprint goals, and recent activity.

Orientation Instructions (`instructions`)

Engrams provides a built-in instructions command designed to output a standardized onboarding instruction manual for LLM agents. You can print this directly to stdout to dynamically brief the agent on how to use engrams in the project:

engrams instructions

This markdown content can be easily piped into your agent setup or saved to your agent instructions (e.g. AGENTS.md or .cursorrules) to teach the agent to run the orientation commands, document decisions, and leverage the knowledge base.

Orienting an AI Agent Instantly

Instead of fanning out multiple commands or copy-pasting massive prompt templates, the recommended way to orient your AI agent at startup is running a single, token-budget-aware command:

engrams prime [--budget <tokens>]

The prime command generates a consolidated briefing containing the product context, active context, recent decisions, patterns, and progress. If a --budget limit is specified, Engrams will automatically drop less critical sections to fit inside your token budget.

Fine-Grained Retrieval Commands

If the agent needs specific pieces of context individually rather than the full consolidated briefing, they can query:

Read the Product Context:
Understand the project's purpose, tech stack, and primary architecture goals.
```
engrams product-context get
```
Read the Active Context:
See what the team is currently working on, open questions, and immediate next steps.
```
engrams active-context get
```
Fetch Recent Activity:
Get a chronological digest of recent changes to see what was logged recently.
```
engrams activity --hours 48
```
Generate a Full Report:
Get a complete snapshot of all knowledge types, or open the interactive HTML dashboard using:
```
engrams report open
```

Example Onboarding Session

Here is what a typical agent startup sequence looks like under the hood:

AI: [Running: engrams prime --budget 1000]
    Result: {
      "product_context": { "stack": "Rust, SQLite", "purpose": "High-performance CLI tool" },
      "active_context": { "focus": "Implementing link commands", "status": "InProgress" },
      "decisions": [
        { "id": 12, "summary": "Use rusqlite with bundled feature", "timestamp": "2026-07-10T12:00:00Z" }
      ]
    }

AI: "Hello! I see this is a Rust CLI tool utilizing SQLite (via rusqlite with bundled features).
    The current active focus is implementing the 'link' commands. I am ready to assist."

Summary of Onboarding Commands

engrams prime — Generates a consolidated, token-budget-aware briefing of all contexts.
engrams product-context get — Retrieves the stable high-level overview.
engrams active-context get — Retrieves the current session focus and status.
engrams activity [--hours <N>] — Generates a chronological digest of recent changes across all memory records.
engrams report — Generates a comprehensive project knowledge report (supports JSON and interactive HTML dashboard via report open).