Contribute

The most valuable contributions to Mind Archive are not code.

Outside code is not being accepted yet

Accepting outside code would need a licensing arrangement that does not exist yet, so pull requests from outside the project cannot be merged. Opening one risks wasting your time, which we would rather avoid.

This is stated plainly rather than left for you to discover after doing the work.

What is genuinely useful right now

Tell us about an export that will not import. Every provider changes its format eventually, and one real broken export is worth more than any amount of guessing. Say which service, when you exported, and what happened.

Never send the file itself. It is a copy of your private conversations. A description of the shape — the top-level keys, roughly how many conversations — is what is actually needed.

Say what confused you. Wording that made you hesitate is a bug. So is a documentation page that assumed something you did not know.

Test it somewhere we cannot. Other operating systems, other shells, other Docker setups, unusual filesystems. “It worked” is useful; “it worked except this” is more useful.

Ask a question. A question that reveals unclear documentation improves the documentation.

When the CLA exists

The work that will be most welcome:

A new importer. Every provider sits behind its own adapter, and the core imports none of them — so a new one is genuinely self-contained. Detection is by the shape of a file rather than its name, because ChatGPT and Claude both export something called conversations.json. See the architecture.

Test fixtures must be synthetic. Test files are committed, and a real export is a copy of someone’s private conversations. Hand-write a small sample that exercises the format’s shape, and keep any real export you are working against in tmp/, which is git-ignored.

Documentation corrections, at any time.

The principles a change has to respect

Local first. It works with no network and no account.
Nothing leaves the machine unless explicitly configured.
User content stays human-readable on disk. SQLite indexes the archive; it never owns it.
No provider sits at the centre.
No dependency that cannot be justified in one sentence.

Writing a blog post

The blog is part of this site, in docs/_posts/. docs/_drafts/TEMPLATE.md carries the house style.

Draft a post however you like — including with an AI assistant. Nothing is published without a person approving it. No automation posts here. If an assistant drafted it, you are still the author, and the accuracy of every claim in it is yours.

A post is not expected with every change. Write one when something is worth explaining to someone outside the project.

The full guide

Everything above is the summary for a site visitor. The canonical document is CONTRIBUTING.md, alongside AGENTS.md — which applies whether you are a human or an AI coding agent.

Security issues should not go in a public issue. See the security notes.