Product / Memory
Context you can come back to.
Keep and edit useful agent memory while treating important facts as reviewable information.
Save the durable context
Preferences, recurring constraints and a project vocabulary are good candidates for memory. A temporary guess or a copied secret is not. State what you want an agent to remember and inspect the saved note before treating it as a lasting instruction.
Separate memory from evidence
Remembering a preference helps an agent tailor its work. It does not turn an old conversation into a current source. Ask for fresh evidence when dates, product details or external conditions matter.
Review what persists
The product includes memory views and editing, with behavior that can vary across engines. Privacy and cross-engine memory qualification remain ongoing. Review a note after changing it, check which agent or room it belongs to, and remove obsolete instructions.
Keep exploring
Your first agent, step by step.
Set up the Suparagent hosted preview with an email address, your sandbox-provider key and your AI account.
Security & dataUnderstand where the work runs.
Learn the boundaries between your browser, the hosted service, sandbox providers and AI engines.
Use casesPut a team around a real task.
Explore practical agent workflows for research, development, content and operations.
Give the work a home
Build your first AI team.
Start with one agent, one clear task, and a result you can review.
Preview · Bring your sandbox provider and AI account