Stop re‑explaining your repo to AI agents. Stop re‑explaining your repo to AI agents.

npm install -g @aictx/memory

Why Memory?

Durable project memory for AI agents: give the repo shared context, load only what matters, and inspect what agents remember.

01 Set it up once.

Give your repo durable memory for the product intent, decisions, workflows, conventions, and gotchas agents usually need you to repeat.

02 Load what matters.

Agents pull focused context before work — no long briefing, giant prompt, vector database, or retrieval stack to maintain.

03 See what agents remember.

Inspect the same typed memory agents use, including objects, schema, relations, provenance, and graph context, in a local visual viewer.

Instead of reopening old decisions, the agent starts with a small, task-shaped context pack before the edit begins.

Can you polish viewer search?

Worked for 7m 18s⌄

I need to inspect the docs and viewer code to recover the product direction.

Ran rg "viewer|search|schema|facets"

Read docs/viewer.md, viewer/src/App.svelte

Explored 11 files, 4 searches

Remind me: are we still avoiding dashboard-style UI here?

Can you polish viewer search?

Ran memory load "improve viewer search UI"

AI Context Pack Viewer memory page shows canonical types, facets, scopes, and relations first.

Keep guided lenses secondary. Verify schema filters and relation boundaries.

Loaded 6 relevant memories before editing

I have the UI constraint and verification path. I’ll update search without reopening old decisions.