Shared context
Shared context is the information that everyone working on something can see: people and their AI agents alike. That includes the brief, the decisions made so far, the relevant files, who’s doing what, and what changed yesterday. It’s the opposite of context that lives in one person’s head or one private chat.
Why it matters
An AI model only knows what is in its context window. If each person loads their own agent with their own copy of the background, the copies drift. One agent works from last week’s spec, another from today’s, and they produce work that doesn’t fit together. Every new conversation also starts with someone re-explaining the project, which is slow and never quite complete.
When the context is shared, a decision written down once reaches every agent that needs it, and a new teammate, or a new agent, starts with the full picture.
Using it well
- Write decisions down where the work happens, not in a meeting nobody recorded.
- Keep it current. Stale shared context is worse than none, because agents treat it as true.
- Keep it scoped. Shared doesn’t mean everything visible to everyone. Each project or team should share what’s relevant to it, with access rules that match.
- Don’t load it all at once. An agent should be able to look things up in the shared material, rather than having all of it pushed into every request.
Shared context vs. memory
Memory belongs to one agent or one person and helps it remember across sessions. Shared context belongs to the team. Both matter, but only shared context keeps several people’s agents working from the same page.
How elvex relates
In elvex, threads, files and to-dos live in Spaces, so people’s agents read the same decisions. An agent can pick up a write-up a teammate’s agent posted and follow it, and flag what it doesn’t cover. Each agent still sees only what its person can.