Agents

Agents in SuperAGI are complex systems that leverage the power of LLMs to operate independently, reason, perform tasks, and takes decisions without the constant need for human intervention.

The Agents, plan, reason, interact with toolkits, and execute based on the user's Goals, Instructions, Resources, and Constraints to achieve the expected results.

To dive deeper into how Agents operate, you can explore the sections below:

📄️ Agent Provisioning

Introduction:

📄️ Agent Details

Agent Details provide you with a complete view of the Agent’s Performance and a summary of the details you had given while provisioning an agent.

📄️ Activity Feed

The agent feed page provides you with complete, real-time visibility of the activities performed by the agent.

📄️ Pause & Resume an Agent

You can pause an agent during the middle of a run of resume a paused agent if you would like the agent to continue operating.

📄️ Agent New Run

Agent New Run provides you with the capability to start a New Run of an Agent that has been previously provisioned.

📄️ Run History

The Agent Run History provides you with a view of the runs that the agent has performed. It shows the last time the agent has run and the number of calls it has made

📄️ Action Console

The Action Console is used when you are running a permission-based agent. When the agent is running, you can provide directions to the agent before it proceeds to execute the next action such as approving the action, denying and denying with feedback.

📄️ Agent Scheduling

Agents can now be scheduled to run at a particular time and also run recursively at pre-defined timer intervals and also after ‘n’ number of run completions.

📄️ Agent Type

Agents in SuperAGI currently follow Re-Act architecture, where there is a thinking step that is performed by the LLM and an Execution with the tool that is suggested in the thinking step.

📄️ Agent Performance Monitoring

APM is the analytics dashboard that gives detailed insights into all the agents, models, tokens, runs, etc.

📄️ Delete Agent

You can delete a provisioned agent - the activity feed and the run history associated with the agent.

📄️ Edit Agent Template

You can make changes to an existing template - such as modifying the goals, instructions, toolkits, the agent type, etc.