Why I built Heald (and why it's called that)

Hey maker here. Heald started from a dumb but constant annoyance: I'd make an architectural decision in one AI agent, switch to another for a different task, and have to re-explain everything from scratch. Same project, same decisions, zero memory carried over.

Every fix I found for this was either a cloud service, a Node app with a web UI, or an MCP server I had to keep alive. I wanted something that just... sat in the project as files. So Heald is a single Rust binary install it, run `heald init`, and it starts writing plain Markdown into `.heald/`. No database, no server, nothing to run in the background.

The name: a heald is the wire on a loom that threads through every warp strand so it stays aligned to the same pattern, controlled by the loom's harness. Felt right for a tool whose only job is keeping every agent reading from the same thread.

What it does today:

- `heald remember` / `heald blame` — decisions become git-diffable Markdown files, and you can trace which one touched a given file

- `heald map` — a cheap directory map so agents stop grepping for structure they could just read

- `heald forget` / `heald doctor --fix` — memory that stays useful instead of accumulating forever

- One `heald sync` compiles your rules into the format every major agent expects (Claude Code, Cursor, Antigravity, Hermes, Codex)

It's MIT licensed, v0.1, and I'd genuinely like to know where it breaks for your setup — especially multi-agent workflows I haven't tested against. Happy to answer anything here.