Caller
01
Open Inference
The first visit mints a key with 5¢ on it. The key lives in this browser — copy it to keep it.
02
Pick a live model
The board shows who is awake. The cheapest offer for that model answers.
03
Call it
OpenAI-shaped. Wait up to 90 seconds. A slow answer stays readable by job id for 10 minutes at GET /api/v1/jobs/{id}.
curl --max-time 90 https://owterminal.com/api/v1/chat/completions \
-H "Authorization: Bearer sk-owt-…" \
-H "Content-Type: application/json" \
-d '{"model":"MODEL","messages":[{"role":"user","content":"Say hi"}]}'Host
01
Install the ow CLI
One line on the machine that sees the GPU. It times one short completion and posts that to the desk.
02
Join, then serve
ow join registers the loaded models. ow serve is the heartbeat — stop it and the machine leaves the board.
03
What you earn
80% of every finished job. Payouts queue after the first dollar and settle in shielded Zcash.
curl -fsSL https://owterminal.com/ow-install.sh | sh export PATH="$HOME/.ow:$PATH" export LITELLM_BASE=http://127.0.0.1:4000 export LITELLM_KEY=sk-local ow join && ow serve