CLI reference

at1 login       --key <API_KEY>     # connect this CLI to your account
at1 whoami                          # show the connected account
at1 logout                          # disconnect (remove ~/.at1/config.json)
at1 compress    <codec> [--stream] [--row-group N] [--backend xz|zstd] IN OUT
at1 decompress  IN OUT
at1 optimize    compress IN OUT [--bound F]   # auto-select codec; never worse than baseline
at1 optimize    decompress IN OUT             # F>0 => Bounded mode (guaranteed max error)
at1 optimize    verify IN                     # integrity + (if bounded) error-bound certificate
at1 query       FILE --where COL:LO:HI --eq COL:VAL --select C0,C1 [--limit N]
at1 sql         FILE "SELECT cols WHERE ..."
at1 info        FILE                 # codec, schema, row-groups, integrity status
at1 arrow       FILE [OUT] [--format ipc|feather]   # export a queryable .at1 to Apache Arrow
at1 integrity   FILE                 # verify the embedded SHA-256 trailer (byte-exact)
at1 verify      <codec> IN
at1 bench       <codec> IN

login / whoami / logout connect this machine to the account you create on the dashboard. Create an API key under API Keys, then at1 login --key <API_KEY> stores it in ~/.at1/config.json (env AT1_LICENSE_KEY / AT1_METER_URLoverride it for CI). Every encode/query is then metered against that account's plan. Creating archives (compress/encode) requires a free connected account; decoding, querying and decompressing are always free and need no account. The one-time 100 GB-compressed free trial is a credit, not a hard cap: once it's used, add a card and encoding simply continues at the metered rate. A deny — card_required (over the free trial with no card on file), payment_required (unpaid / past-due), or update_required — aborts the operation with a one-line error: usage blocked: … rather than writing output. See Free tier & limits.

info / integrity / arrowwere added for inspection, tamper-evidence, and the Arrow bridge into every engine — and they're the tools the AT-1 MCP server (mcp_server/ in the repo) exposes to AI agents (Claude, Cursor, ChatGPT).

Codecs

Tools

at1_adapter.py    put/get/tier/untier   # transparent storage tier
connectors/at1_ingest.py                # export -> queryable archive
at1_logquery.py   build/query           # queryable-by-time logs
make_proposal.py                        # one-page PDF savings proposal