I work with AI agents every day (coding agents, chat assistants) across several client projects. What costs me the most isn't the work itself, it's the context: re-explaining past decisions, rules that apply to one client but not another, remembering what an agent did last week and why. And switching tools makes it worse, since each one keeps its own memory.
I'm curious how others actually handle this:
1. When you start a new session or conversation, how does the agent get what it needs? Instruction files, the tool's built-in memory, a notes folder, an MCP server, copy-pasting, something you built yourself?
2. Do you use more than one agent or assistant (Claude, ChatGPT, Copilot, Cursor, Codex, Gemini...)? How do you carry context from one to another?
3. If you work for several clients or teams: how do you keep their context separate? Has it ever mixed or leaked?
4. The last time it went wrong (an agent redid a rejected approach, forgot a decision...), what happened, and how much time did it cost you?
5. Have you ever paid for or built a tool for this? Do you still use it?
Disclosure: I'm exploring whether to build something in this space, so I'm trying to learn how people really work, not pitching anything. There's no link. Happy to continue here or by email, my address is in my profile.
Comments URL: https://news.ycombinator.com/item?id=50403153
Points: 1
# Comments: 0
