FEATURES / SELF-IMPROVEMENT
Agents that learn as you work with them
Every interaction makes your agent better. They remember your preferences, learn from corrections, and carry that knowledge into future work automatically.
Every interaction makes your agent better. They remember your preferences, learn from corrections, and carry that knowledge into future work automatically.
Corrections, preferences, and decisions carry forward into future work.
Tell your agent what to do, just like you would a teammate.
Review their work and teach them your preferences as you go.
Your agent remembers what it learned and applies it next time.
Your agent remembers feedback and adapts to your workflows.
A chatbot forgets when the conversation ends. Every Skydive agent has durable, file-based memory that persists across conversations. Tell it a preference or correct it once and it sticks, so it gets sharper the more you work with it. You never re-explain the basics.
No. The agent writes to its own memory as it works. When you state a durable preference, correct it, or establish a fact it will need again, it records it. There is nothing to configure and no notes to maintain.
Yes. Memory is agent-level, not per-person. When one person teaches the agent something, everyone who talks to it next benefits, like a team wiki the agent maintains for itself. Keep private or sensitive details out of a shared agent for that reason.
Yes. Memory is editable. Tell the agent to update or forget something and it will. Change an address, a policy, or a preference, and the agent stays current from then on.
Yes. Memory follows the agent, not the channel. Teach it something in Slack and it remembers in email and web chat too, because it is one agent with one memory wherever you reach it.
Preferences, roles, and working styles for people; context, decisions, and status for projects; general lessons and corrections; and how your external systems and accounts are set up. It organizes all of it by subject so it recalls the right thing in the right context.