My work — prompt any CLI you have, from one page
Write a prompt, pick which AI CLI answers it, and read the reply on the same page. Send the same prompt to a second tool and both answers sit side by side. Everything runs under your own login on your own machine.
How to do it
- Open My work. Every tool found on this machine is listed under Send with.
- Write your prompt — or press Generate prompt to sharpen it first.
- Pick a tool and press Send. The answer appears below, with how long it took.
- Pick a different tool and send again to compare — the first answer is kept.
Two readings of the same question
Send "why is this migration slow?" to Claude Code and to Codex. Both answers stay on the prompt, so you can see where they agree and where only one of them spotted something.
The tool you pick is the tool that runs
There is no silent substitution here. Asking what Codex makes of something is the whole question, so if Codex cannot answer you are told why — not handed a different tool's reply with a different tool's name on it. Prompt generation is the opposite case and works the other way: there, any tool that can write you a good prompt will do, so they are tried in turn.
Your prompt is sent exactly as you wrote it
Nothing is wrapped around it and nothing is added to it. An instruction of ours folded into your words would mean the answer belongs to a question you never asked.
A tool that cannot answer says so
Not signed in, wrong plan, an outage: the failure is kept with the prompt in the tool's own words, and the prompt is saved either way. A tool that failed recently still appears in the picker, marked, because only you know whether you have just fixed it.
Where it is kept
On your machine. Prompts and answers live in your own data directory and stay there unless you ask otherwise. Sync chat history sends the whole history — each prompt and the answer each tool gave — to your account, where it sits beside your prompt timeline and one search covers both. Nothing leaves until you press it.
If nothing is installed
LLMLinq runs the tools you already have and installs nothing on your behalf. With no CLI signed in, Ollama runs an open model locally and free — weaker than a real CLI, and honest about it.