AI Total Recall is a memory cockpit for Claude Code.
Your assistant has a memory now. You have never been able to see it. This mod puts it on screen: what it recalled, what it learned, and the rules it works under, live, in a pane beside your work. Search memory without spending a token. Behind it is CogmemAi: the top published score on LongMemEval, plus guardrails, scope and review that no other memory offers.
/plugin install aitotalrecall --marketplace hifriendbot/aitotalrecall

One fix on a payments project, start to finish: the recall before the work, the pane, a zero-token search, the rules in force, and a save landing with a toast.
No sound? The narration starts at once. If the player will not start, open the MP4 directly.
Most Claude Code mods are gauges: context, tokens, files. AI Total Recall is the first that shows what the assistant actually knows about you and your project.

The pane lists each memory the assistant looked up this session: subject, importance, scope and the first plain sentence. Press Enter on a row to expand it, copy it, start a prompt about it, or find related memories.

The search box talks to your memory server directly. No model turn, no tokens, results in about a second. Also /recall find <query> from the prompt.

Every memory call in the transcript draws as a readable row with the subjects beneath it. The band above the prompt carries live counts, a per-turn activity pulse, and the last thing that happened.

The Rules tab lists the hard rules the assistant works under, one line each. When it learns something new, you see it saved, with a toast, the moment it lands.
In a Claude Code terminal session. Answer y to add the marketplace, then pick the user scope.
/plugin install aitotalrecall --marketplace hifriendbot/aitotalrecall
AI Total Recall watches the CogmemAi memory tools. If none is connected, the pane says so and gives you the setup line. The free tier is enough to start.
npx cogmemai-mcp@latest setup
Type /recall, or click Open on the band. r, l, u switch tabs. Tab walks rows, Enter expands, Esc returns to the prompt.
/recall /recall find idempotency key /recall hide
When the assistant recalls the wrong thing, you find out after the code is written. When it forgets a rule, you find out in production. AI Total Recall makes memory a thing you watch, not a thing you hope.
See the memories that shaped an answer before you accept it. A stale decision or the wrong project is one glance away.
Hard rules load silently into every session. The Rules tab shows exactly which ones are in force right now.
Search memory yourself, copy a memory into a prompt, or ask about one. No tokens spent until you decide to spend them.
Most Ai memory is a vector store with a save button. CogmemAi is the engine behind AI Total Recall, and it does four jobs the others do not: it remembers with the best published recall in the field, it enforces your rules before an action runs, it knows what each project is for, and it reviews the work when the turn ends.
The mod is where it starts. The rest of the line runs on the same CogmemAi engine, so the memory you see in Claude Code is the memory that follows you to every other Ai you use.
The memory behind the pane is measured in public, with the scores, methodology, and dates to back it.
A free mod for Claude Code. It adds a pane beside your work that shows what the assistant recalled this session, what it saved, and the hard rules in force, with a search box that queries your memory server directly. It is powered by the CogmemAi memory engine.
Yes. The mod is free and MIT licensed, with the source on GitHub. It needs a CogmemAi memory server to watch, and CogmemAi has a free tier.
No. The mod runs inside your Claude Code session and makes no network calls of its own. It reads the results of memory tool calls that are already happening in your session, and a pane search calls the same memory tool the assistant would. Nothing leaves your machine that was not leaving already.
Claude Code 2.1.296 or newer, where mods (plugins of function hooks) are supported. It draws on the terminal and the desktop app.
The CogmemAi memory tools (recall, preflight, project context, and the three save tools), the prompt, the pane and band drawing sites, and two slash commands. It never changes a tool result on the way through. The full list is in the README, and the tests run with claude plugin test.
Because it is the smartest memory you can plug in. It holds the top published scores on LongMemEval (95.1%) and LoCoMo (91%), and it does things a plain memory store cannot: Guard enforces your hard rules before a command runs, Scope gives each project an intent document so recall stays relevant, Review grades the work against that intent at the end of a turn, and corrections are saved with the right approach so the same mistake does not repeat. AI Total Recall is the window onto all of that.
Not yet. Today it watches the CogmemAi tools. Adapters for other servers are on the roadmap; the repo is open if you want to help.
HiFriendbot's family of memory products, all on the CogmemAi engine: this mod, the engine itself for developers, Cognitive Memory for chat assistants, CogmemAi for Teams, A Brain for Robots, CogmemAi for Frontier agents, and drop-in memory for Claude Code, Cursor, Windsurf, and ChatGPT.
One line in any Claude Code terminal. Free. The memory behind it has the top published score on LongMemEval.
/plugin install aitotalrecall --marketplace hifriendbot/aitotalrecall