read the
damn manual.
an AI shell helper that gives you the command. nothing else.
Stop pasting man pages into ChatGPT. rtdm lives in your shell and gives back commands, not essays.
01three modes
One tool. Three flags.
Three things, one keystroke each. No subcommands, no config.
-
(default)
Get the command.Describe what you want. Get the command back.
-
-e, --explain
Explain a command.Paste a command, get a plain-English breakdown of every flag.
-
-q, --ask
Ask a question.For when you want an answer, not a command.
02under the hood
How it actually works.
Hosted inference, open-source CLI, no third parties.
-
GPU
Runs on a consumer GPU.Open-weight coding model on dedicated hardware. ~200ms once warm.
-
CLI
Open source, MIT licensed.Whole client is on GitHub. Don't trust our endpoint? Point it at your own Ollama.
-
NET
No third-party AI in the path.No proxying to OpenAI or Anthropic. Inference happens here. Queries aren't logged or trained on.
03pricing
Two ways in.
$3/mo hosted, or self-host with Ollama. Same code either way.
500 queries per month. Hosted on our GPU. Nothing to install beyond the CLI.
Cancel anytime. Use through your billing period. No refunds.
Run it locally with your own Ollama. Same client, points at localhost:11434.
MIT licensed. Bring any model Ollama supports.
04privacy
A note on privacy.
No query logging. No training on your data. No analytics exception. The request comes in, the response goes out, the request is gone.
One caveat: Cloudflare sits in front for DDoS protection, so they technically see TLS-decrypted traffic. Most of the web works this way, but worth saying out loud.
Want zero third parties? rtdm is open source. Run your own Ollama and never touch our servers.