Instructions
Tell every agent once.
Instructions are the rules you write yourself, in Settings. Every Claude Code and Codex request on every machine carries them.
// Settings → Instructions · 4,096 characters at most
Keep answers short. No wall of text.
Never touch code or technical terms.Saved once. Every request from every agent carries it, read from the cache.
Instructions: Explain each change in detail. Your message: Answer in one line. The agent: one line.
Instructions are a standing rule. What you ask now comes first.
| Fact | Value | Valid from | Until |
|---|---|---|---|
production deployed_to | Hetzner Cloud, Falkenstein | 21:14:38 | 21:15:09 |
production deployed_to | Render, Oregon region | 21:15:09 | now |
Memory is learned from your sessions. Instructions change only when you save.
You set the rule. Every agent follows it.
- 01You write it
“Keep answers short.” Up to 4,096 characters, in Settings.
- 02Every request carries it
Claude Code and Codex, on every machine. No file to copy into each repo.
- 03Your message still wins
Ask for detail in one turn, and you get detail.
You write one. The other is learned.
Written by you. Change only when you save. How the agent should answer you.
Learned from your sessions. Changes as facts change. What is true about you and your projects.
Costs almost nothing.
After the first request, it is read at the cache price, not as new input.
It goes at the start of the conversation. Your client stays as it is.
Both agents, every machine. Clear the box to turn it off.
Write it once.
npx @context-mode/cli connects your agents. 1,000 requests free, no card.