Guide
What are AI agents?
Understand how an AI agent combines instructions, tools and context to work toward a result.
From an answer to a task
A chat model generates a response from the context it receives. An agent workflow adds a way to take steps: inspect a file, run a tool, evaluate the result and decide what to do next. “Agent” describes that workflow; it does not guarantee autonomy or accuracy.
The building blocks
Useful agents have a clear goal, access to the right material, tools appropriate to the task and a stopping condition. Persistent instructions and memory can carry context forward. Approval controls help a person review consequential steps.
A concrete example
For a research brief, the agent reads provided documents, extracts claims and sources, writes a draft and returns unresolved questions. You check whether the sources support the conclusions. The useful output is a traceable brief, not simply a confident answer.
Keep exploring
A practical guide to working with agents.
Learn the basics of AI agents, set up Suparagent and build reviewable workflows.
ProductGive every agent a place on the team.
Explore named agents, teams, memory, routines, skills and review controls in Suparagent.
Security & dataUnderstand where the work runs.
Learn the boundaries between your browser, the hosted service, sandbox providers and AI engines.
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