Briefing

Before you direct the work, know the state of the work. Three commands — a token-budgeted brief, a full dashboard, and a health audit — mean you never start a session cold.

engrams prime — the session brief

What you run:

engrams prime
engrams prime --budget 4000
engrams prime --paths src/graph,src/ontology --tags policy-engine

What you get: a single context brief sized to a token budget — the active context, the decisions and patterns that matter right now, and recent progress — ranked and packed until the budget runs out. --budget sets the approximate token ceiling, --paths narrows the brief to entries anchored to specific files, and --tags filters by topic.

Why it's valuable: this is the command that turns a blank-slate agent into one that starts the session already oriented. On a young graph prime returns a short reading list. On a mature one it returns the reasoning behind the code you're about to touch — which decisions are still active, which were superseded, and which standards apply to those paths. Your agent runs this first; you skim the same output to see exactly what it was told.

engrams report — the project at a glance

What you run:

engrams report                 # full structured report
engrams report decisions       # one topic: context | progress | decisions | patterns | links
engrams report open            # HTML dashboard in your browser
engrams report open --no-browser --out ./report.html

What you get: a structured report of everything the project knows about itself, with --limit capping each section (default 50). The report open variant generates a browser-based HTML dashboard — including an interactive view of the knowledge graph — and opens it for you, or writes it to --out with --no-browser for sharing.

Why it's valuable: the CLI output is for the agent; the dashboard is for you. One command renders the entire state of the project — what was decided, what's in flight, how the pieces connect — as a page you can scan in a minute over coffee. It's the fastest way to answer "where is this project, actually?" without reading a single commit.

engrams doctor — the health audit

What you run:

engrams doctor

What you get: an audit of the database itself: orphaned records, drift between what's stored and what's true, missing context, and non-canonical relationship vocabulary that should be folded into the ontology.

Why it's valuable: an advisor is only as good as its record. A stale decision that was never marked superseded, or a pattern with no anchors, silently degrades every brief and every check built on top of it. doctor catches those problems early — while they're a lint warning, not after they've poisoned a session's advice. Run it like you'd run a test suite: regularly, and definitely before you trust a surprising answer.

Keep exploring

• For You — the human's relationship with the advisor
• Enforcement — turn standards into checks that run before violations land
• Exploration — interrogate the graph directly
• The Advisor — why briefing gets better the longer engrams runs