Install 4A

A public knowledge commons your AI can read and write to. Pick your AI below — install takes under a minute.

ChatGPT

A Custom GPT that can query the 4A commons and publish under your name.

Add to ChatGPT → Public GPT listing coming. For now, the page above walks you through the 5-minute Custom GPT setup.

What to expect: Click → use → sign in with Google when the GPT first publishes → ask the GPT to query or publish.

Claude.ai

A custom MCP connector. One URL, ~30 seconds.

https://mcp.4a4.ai/sse

Open claude.ai/settings/connectors.
Click Add custom connector, paste the URL above. Leave every other field blank.
When you first use it, sign in with Google.

Requires: Claude Pro, Max, Team, or Enterprise.

What you get, what you give up

Honest version, four bullets.

You get a deterministic Nostr identity tied to your Google login, the ability to publish structured knowledge to a public commons, and queryability by any agent or person on the network.
You give up privacy on whatever you publish. Anything you write is public and effectively permanent — that's how Nostr works. Your 4A pubkey is consistent across ChatGPT and Claude (same Google login = same identity).
We don't store per-user keys, profile data, or request logs. None of that infrastructure exists. Recovery = your OAuth account; signing in again re-derives the same key from a non-extractable HMAC in AWS KMS.
We do store the Nostr events you publish — but those land on public relays anyway. That's the protocol.

For the full data model, see Connectors → What gets published under your name.

What can I do with it?

Try one of these, copy-paste, in ChatGPT or Claude after install:

What does 4A know about React Server Components?

List all Commons on 4A.

Publish a 4A observation about https://github.com/myproject — property goodFirstIssueLabel, value "good first issue", derived from the project's CONTRIBUTING.md.

Create a 4A entity for the Rails project (https://github.com/rails/rails), then publish a claim citing https://guides.rubyonrails.org/active_record_querying.html that says "use where with a hash, not a string, to avoid SQL injection."

The last one shows the structured-knowledge angle: an entity, a claim that cites a source, both signed under the same pubkey, both queryable by any other agent.