A Chrome extension that shows what an Agent Memory backend is holding for a given user - working memory on the left, long-term memory on the right, both auto-refreshing.
- DevRel & solutions engineers - drop it onto any demo to show what's happening in memory live; no need to build a custom viewer or memory panel for every new demo or talk you create.
- Teams building agent apps - debug memory issues during development without.
- Workshops & tutorials - give learners a clear window into how extraction, search, and retention actually behave.
Works with either:
- Redis Agent Memory Server - the open-source server you self-host (e.g. via Docker)
- Redis Agent Memory - the hosted service on Redis Cloud (Iris / Context Engine)
Pick the backend from the connect panel; the rest of the UI is identical.
| Pane | Scope | What it shows |
|---|---|---|
| Working memory (left) | Current session_id only |
Message log, role tags, per-message discrete_memory_extracted flag, the running summary if AMS has generated one |
| Long-term memory (right) | All sessions for user_id (+ optional namespace) |
Extracted memories, type badges (semantic / episodic / message), topic + entity chips, originating session_id, similarity score when search is active |
chrome-extension/ the unpacked extension Chrome loads
proxy/ optional transparent proxy for Cloud (see "Known issues")
- Open
chrome://extensions. - Toggle Developer mode on.
- Load unpacked and pick the
chrome-extension/folder. - Pin the extension to the toolbar.
Click the toolbar icon to open the connect panel:
- Backend - pick OSS server (Redis Agent Memory Server) or Redis Cloud (Redis Agent Memory service)
- Redis Agent Memory Server URL / Endpoint - e.g.
http://localhost:8000for OSS, or the Cloud endpoint (e.ghttps://gcp-us-east4.memory.redis.io) - Store ID - Cloud only; the store identifier from the Redis Cloud console
- API Key - Cloud only; the bearer token from the Redis Cloud console
- Proxy URL - Cloud only, optional; override the built-in default
- Namespace - OSS only, appears if the server has any
- User ID - optional combobox
- Session ID - combobox of sessions for the chosen user / namespace
- Working / Long-term memory refresh - in seconds (default 3s and 5s)
Click Connect.
| Action | Where | Effect |
|---|---|---|
| Search long-term memory | Search box top of the right pane | Hybrid (vector + keyword) search via /v1/long-term-memory/search. Score pill appears on each card. |
Toggle optimize_query |
Checkbox next to the search box | LLM-rewrites the query server-side before searching |
| Filter by topic / entity | Click any chip inside a card | Adds to the active filter; pills above the cards let you remove or clear |
| Delete a long-term memory | Hover any card → ✕ top-right |
DELETE /v1/long-term-memory?memory_ids=…, confirms first |
| Clear working memory | Clear button in the working pane header |
DELETE /v1/working-memory/{session_id}, confirms first |
| Refresh now | ↻ in the header |
Immediate poll of both panes |
| Reconfigure | Reconfigure link in the header |
Back to the connect form |
Direct fetches to *.memory.redis.io are blocked by Cloudflare bot detection. Route through the included proxy/ - cd proxy && npm start runs it locally on :8787, or deploy it to any Web-fetch runtime - and put its URL in the Proxy URL field of the connect panel.
MIT.

