an AI chat client
teemoon has no
server.
the same models, on your key or your phone — nothing touches a teemoon machine.
scan ona phone
the safer path existed all along.
call the same model through its API and it doesn't train on you by default. teemoon is built for that route — or for a model that never leaves your phone, or one on hardware you own.
run one on your phone, point it at your own machine, or paste in a cloud key.
who gets to read it?
end-to-end on near.ai’s own fleet. everywhere else, plain tls.
- encrypted all the way to the modelciphertext leaves your device; only the attested enclave ever sees plaintext.
- plaintext exfiltration, auditedthe attested images ship with a fable miniaudit asking one thing: does that code copy your message anywhere besides the model? these are teemoon’s own ai-assisted reads — not an outside firm’s.
- anchored in siliconintel’s root of trust on the device; nvidia’s attestation service over tls. a failed check stops the send cold; an inconclusive one asks you first.
written up by near devhub, 18 aug 2026 · reviews at teemoonai/audits
settings → developer mode · a debug card follows every reply
where your prompts actually go.
all research →-
pay per token, not per month — but the cheap-model era has a plot twist.
metered billing beats a $20 subscription for ordinary use and loses badly at the heavy tail. the arithmetic, the vendor-vs-reseller distinction, and why the cheap-open-weight discount is compressing.
read it → -
your AI chatbot trains on you. the api running the same model doesn't.
consumer tiers train on your conversations. api tiers don't, and they cap retention where consumer tiers don't cap it at all — same model, same prompt, different defaults. then the part the marketing skips: the api isn't private in ways that matter either.
read it →