Quickstart
The fast way: hand it to your AI
Paste this to your coding AI (Claude Code, Cursor, …) — it clones, installs, and indexes in one go:
Install whatever-recall from https://github.com/heidrich/whatever-recall:
clone it, `pip install -e .`, then run `recall init .` in my repo and
`recall explain` to orient. From now on, run `recall brief <file>` before
editing any file.
That's it — your AI does the rest, and from the first commit your code starts remembering its own reasons.
The manual way
recall is a Python package — install it from the clone (or the wheel once published):
git clone https://github.com/heidrich/whatever-recall
cd whatever-recall
pip install -e .
# or, once published:
pip install whatever-recall
On Windows, the CLI expects UTF-8 output:
set PYTHONIOENCODING=utf-8 # cmd
$env:PYTHONIOENCODING = "utf-8" # PowerShell
export PYTHONIOENCODING=utf-8 # bash
Index your repo
From the repo root:
recall init .
This builds the .mind/ index (a local SQLite database, git-ignored): the code map (tree-sitter), the commit history, and any decisions/lessons it can read. It costs 0 model tokens — it's a parser and a database, not an LLM.
Wake it up at session start
recall explain # orientation: load-bearing files, decisions, what's in flight
recall dashboard # opens the browsable brain + graph in your browser
recall hook --install # the pre-commit risk-warning + post-commit auto-stamp
The everyday loop
recall brief src/server/orgs.ts # BEFORE you edit a file: why, what breaks, open tasks
recall "where is the seat ceiling enforced" # ask by concept
recall resolve seatLimit # correct a guessed name into this repo's real vocabulary
recall review # before committing: what this change can break
All of those are read-only and cost 0 model tokens.
Stamp what you learn
After a deliberate decision or a tricky fix:
recall stamp "we refuse a plan downgrade below occupied seats — it would strand members" \
--anchors src/server/orgs.ts
It surfaces automatically in the next brief on that file — so the next session (yours or a teammate's) inherits the reason.
Next
- Core commands — every command, what it returns, when to reach for it.
- How it works — write-time vs read-time, the 0-token read path.