airlock
How it works

Three calls. That's the whole API.

No SDK, no client library, no lock-in. Every operation is a plain HTTP request with headers.

  1. Create a secret

    Send a name, a value and your access key. It's encrypted with Fernet before it ever touches disk.

    $ curl -X POST .../api/create-secret/ \
        -H "access-key: $ACCESS_KEY" \
        -H "secret-name: DATABASE_PASSWORD" \
        -H "secret-value: ..."
  2. Retrieve it, from anywhere

    Any service, any language. It's just HTTP headers, so there's no SDK to depend on.

    $ curl .../api/get-secret/ \
        -H "access-key: $ACCESS_KEY" \
        -H "secret-name: DATABASE_PASSWORD"
  3. Update or delete it

    Same shape, different verb: PUT to rotate a value, DELETE to remove it.

    $ curl -X PUT .../api/update-secret/ \
        -H "access-key: $ACCESS_KEY" \
        -H "secret-name: DATABASE_PASSWORD" \
        -H "secret-value: ..."
Under the hood

Every request is checked, scoped and logged

1. Rate limit

Per-IP limits reject abusive traffic before it reaches authentication.

2. Authenticate

The access key is validated with an O(1) hash lookup, then checked against its IP allowlist.

3. Execute

The secret is encrypted or decrypted with Fernet, scoped to the key's owner.

4. Audit

The outcome, success or failure, is written to the audit log.

See the full feature set

Encryption, allowlisting, rate limits and observability, all in the core service.