It warns you first
You're about to hand-rollback a migration. Engrams surfaces decision #6 —
"use the migrate command, never manual SQL" — before you repeat a
mistake it already watched you make.
Engrams anchors every decision, pattern, and trade-off to your code as a living knowledge graph. On day one it fetches. By week four it advises — warning you before you repeat a mistake, surfacing the rationale behind any line, and telling your agent what breaks if it touches this.
Without memory, your AI repeats mistakes you already fixed, ignores conventions you already decided, and wastes your time re-explaining context it should already have.
Every AI conversation starts cold — you re-explain your project, decisions, and conventions every time
One command rehydrates the full context your agent needs — decisions, patterns, open threads, what breaks if you touch this file
Important architectural decisions and trivial notes rank identically — the AI wades through noise
Important, recent, frequently-consulted decisions surface first via retrieval scoring
Old obsolete decisions pollute context forever
Memory self-cleans: unused records fade into the archive automatically
You hope the AI reads your conventions. It usually doesn't
The system surfaces constraints automatically when the AI edits a relevant file
Convention violations land in your codebase — caught in review if you're lucky
Violations are blocked at edit time and commit time — they never land
Query tools wait for you to ask the right question. Engrams builds enough context to raise its hand first.
You're about to hand-rollback a migration. Engrams surfaces decision #6 —
"use the migrate command, never manual SQL" — before you repeat a
mistake it already watched you make.
Any line of code traces back to the trade-off that created it. Ask why the graph uses PageRank and get the rationale, the rejected alternatives, and the incident that forced it.
Every refines and depends_on link is a tripwire. Touch a
load-bearing decision and engrams tells your agent exactly what else moves.
Your agent does the remembering. You get the briefing.
$ engrams prime
{"briefing": "3 active threads, 1 blocked…",
"relevant_decisions": [ … ]}prime — the morning briefingOne command rehydrates everything relevant to what you're about to touch. Start every session informed, not cold.
Briefing commands →$ engrams report
→ dashboard.html (open in browser)report — see the whole projectA static HTML dashboard of decisions, open threads, and graph health. The state of the work at a glance.
See a report →$ engrams check
✗ src/ops: pattern "no-throwaway-collect" violatedcheck — enforce what you decidedTurn logged patterns into checks that run in CI. The standards you agreed on actually hold.
Enforcement →Every node is a real decision or task from building this tool. Gold nodes are decisions; the bright ones are the load-bearing few everything else depends on.
And this is just the knowledge layer — the full graph, with code topology and git-derived links, spans 802 edges.
Your agent logs each decision and pattern as it works — no extra effort from you.
It connects new knowledge to old: refines, supersedes, depends_on.
prime and relevant push the right context to the front at the right moment.
You and your agent decide with full memory. The graph gets denser. Repeat.
cargo install engrams-clior brew install stevebrownlee/tap/engrams