One memory for all your AI tools.

What you worked out in ChatGPT, Claude Code already knows. Anamnesis collects your conversations from every AI tool into one private memory that all of them can search.

curl -LsSf https://devhaitham481.github.io/anamnesis/install.sh | sh

macOS & Linux · open source · no account · what this does

$ anam sync
✓ 4 tools · 212 conversations · 18,406 messages

$ claude
> why did we pick session cookies over JWT?

● search_memory("session cookies vs JWT")
  ChatGPT · api-server · Mar 4
  "…we need instant revocation, so server-side
   sessions; JWT would mean a denylist anyway."

You decided this in ChatGPT on March 4: you
needed instant revocation, and JWTs would have
required a denylist anyway.

Synced automatically

Claude Code
Codex CLI
Gemini CLI
Cursor

Live, through your node

claude.ai
ChatGPT
their desktop & mobile apps

Imported from exports

ChatGPT
claude.ai
Gemini

How it works

It is only the memory.

Anamnesis never calls a model and isn't another chat app. You keep using the tools you already use; they just stop forgetting.

Collect

Reads the transcripts your AI tools already keep on disk, in the background, picking up only what changed.

Recall

Every agent gets an MCP server with search_memory: keyword and semantic search, run locally, filterable by tool or project.

Remember

Agents save decisions with remember. A note written in Cursor is there the next time you open Claude Code.

Install

One command, then restart your tools.

Nothing to set up beforehand. The installer brings its own Python and never touches your system's.

curl -LsSf https://devhaitham481.github.io/anamnesis/install.sh | sh

Run the installer

It installs uv if needed, then Anamnesis in its own environment. Read the script.

Answer one question

Setup finds your AI tools, imports their history, and asks before connecting them. Every config it changes is backed up first.

Restart your AI tools

Then ask about earlier work: “what did we decide about auth?” New conversations keep syncing on their own.

Rather not pipe a script into your shell? uv tool install git+https://github.com/devHaitham481/anamnesis then anam setup. To update, run the installer again. Windows: use WSL. Uninstall and troubleshooting.

Web & mobile

Your memory in claude.ai and ChatGPT, even while your laptop sleeps.

Run a small node: your own always-on copy of the memory, on a server you control. Web and mobile apps reach it as a connector, and it keeps all your computers in sync.

Your computersClaude Code, Codex, Cursor, Gemini CLI

end-to-end
encrypted sync

Your nodeFly.io, a VPS, a NAS, a home server

MCP connector
with sign-in

Web & mobileclaude.ai and ChatGPT, on any device

Deploy it once

One command on Fly.io, or Docker anywhere. Prebuilt image, HTTPS included.

anam node deploy fly

or, on your own server

docker compose up -d

Link your computers

Enter the node password and choose what to upload: everything, all but some projects, or only notes.

anam node link https://your-node

Add the connector

Paste your-node/mcp into claude.ai or ChatGPT and sign in. Choose search-only or search and save.

anam connect claude-web

Custom connectors need a paid claude.ai or ChatGPT plan. Only need your computers in sync? A folder you already sync (iCloud, Dropbox, Syncthing) works too, with anam share.

Privacy

Yours, on hardware you choose.

There is no Anamnesis cloud and no account. Your memory lives in ~/.anamnesis and, if you run one, on your node.

Local by default

SQLite plus a local embedding model. No API keys, no telemetry. The only download is the embedding model on first run.

End-to-end encrypted sync

Sync data is encrypted with AES-256-GCM before it leaves your computer, with a key only your devices and your node hold.

Apps never see your password

Web apps get revocable OAuth tokens. Give them search-only or search-and-save access.

See every access

anam node activity lists each search and save an app made. anam node revoke cuts one off.

Commands

Everything from one CLI.

Plus a terminal UI (anam) and a web UI (anam-web) to browse and search everything.

`anam setup`	Find your AI tools, import their history, connect them. Safe to re-run.
`anam sync`	Import new history from your local AI tools
`anam connect [tool]`	Give Claude Code, Claude Desktop, Cursor, Gemini CLI or Codex access. Configs are backed up first.
`anam search <query>`	Search memory from the terminal
`anam import <files>`	Import ChatGPT, claude.ai or Gemini data exports
`anam node …`	`deploy`, `link`, `scope`, `status`, `activity`, `revoke`, `wipe`
`anam status`	What is stored and which tools are connected

FAQ

Questions

Does my data go anywhere?

Not unless you set up sync. Without a node, everything stays on your machine. With a node, it goes to your own server, encrypted in transit. The node must read your memory to search it, which is why it runs on hardware you control.

What does a node cost?

About one small machine. On Fly.io the default is 1 shared CPU, 1 GB of memory and a 1 GB volume; it stops when idle and wakes on the next request. A VPS or home server you already have costs nothing extra.

What about Gemini on the web?

Gemini doesn't support custom connectors yet, so it can't read the memory live. You can still import its history from Google Takeout with anam import.

How do agents know when to use it?

The MCP server describes its tools to them. Refer to earlier work ("what did we decide about auth?") and they search. They can also save decisions with remember, or you can ask them to.

Can I keep some projects out?

Yes. Each computer chooses what it uploads to the node: everything, everything except listed project folders, or only saved notes. Change it any time with anam node scope.

Does search need a GPU or an API key?

No. Semantic search uses a small embedding model (bge-small) that runs on the CPU. Keyword search works immediately while embeddings are built.

Stop re-explaining your project.

curl -LsSf https://devhaitham481.github.io/anamnesis/install.sh | sh