attograd memory · a riverfront productAttograd Memory keeps the decisions — we use SQLite, auth lives in this file, do not touch billing — in one file next to your code. Tomorrow's chat starts already knowing them.
free · MITone file, in your repopython 3.11 or newer
6
hosts it installs into
Cursor, Claude Code, Codex, Gemini CLI, OpenCode, and anything else that runs a local tool.
1
file, kept in your repo
No server to run, no account to create, nothing uploaded anywhere.
$0
to use it, ever
MIT licensed. Python 3.11 or newer.
Nothing about how you work. Only what the agent knows when it starts.
A tool you install in the coding agent you already use. You do not manage a database — the agent reads before it starts, and writes when something should outlive the chat.
When you and the agent settle something that should last — we use SQLite, auth lives in this file, do not touch the billing service — Attograd writes that down. Not the whole chat. The parts you would be annoyed to explain again.
Before the agent plans or edits, it reads what this project already decided. It is less likely to undo yesterday's work, or to ask you the same questions twice.
Each project gets its own memory file on your machine, next to the code. Nothing is uploaded and there is no server to run. Close the laptop; the context is still there when you open it.
One install per machine is enough. Pick the host you already work in.
attograd-memory
Install the plugin, choose global scope, then start a new agent chat.
/plugin marketplace add gouthamk16/AttogradDB
Then /plugin install attograd-memory@attograd-plugins.
pip install attogradDB && attograddb setup
Setup detects the tools on your machine and writes each one's config.
pip install attogradDB
The same package, if you want to store and search notes from a script.
The same store without an agent in the middle — notes in, search out, still just a file on disk:
from attogradDB import VectorStore
store = VectorStore(path="memory.db")
store.add(["retry logic lives in client.py"], project="payments")
for doc_id, score, text in store.search(
"where is retry handled", project="payments"
):
print(score, text)
store.close()Free, MIT, Python 3.11 or newer. The consultancy and the audits are a different door — riverfront ai labs if you wanted those instead.