Keep your memory in your repo

Some teams want agent memory to be a file in the repository: readable without any service, reviewable in a diff, and still there if you stop using Hivemind. sync writes that copy.

npx @get-hivemind/cli sync

Run it inside a repository. It reads the memories saved for that project and writes them into one block of AGENTS.md:

<!-- hivemind:start -->
## Hivemind memory for payments-api

- Deploys are capped at three a month, on Tuesdays _(Claude Code, 2026-10-03)_
<!-- hivemind:end -->

What it touches

Only the text between the two markers. Everything else in the file stays as you wrote it, and running sync again replaces the block instead of adding a second one. If the file does not exist it is created.

Memories that were replaced by newer ones are left out. The newest sixty are shown, and the block says how many older ones are not.

It asks first

The file may be committed, so sync shows how many memories it will write and asks before writing. Commit it only if everyone who can read the repository may read them. Add --yes to skip the question, or --dry-run to see the block without writing anything.

Other files

npx @get-hivemind/cli sync --file CLAUDE.md

Importing again

import skips the block, so syncing and importing never feed each other.