Developing AgentKit applications locally is a breeze when combined with the Inngest Dev Server.The Inngest Dev Server is a local development tool that provides live traces and logs for your AgentKit applications, providing a
quicker feedback loop and full visibility into your AgentKit’s state and Agent LLM calls:
The Inngest Dev Server needs to be able to trigger your AgentKit network over HTTP.
If your AgentKit network runs as a CLI, a few lines changes will make it available over HTTP:
You can trigger your AgentKit network by clicking on the “Trigger” button in the Inngest Dev Server from the “Functions” tab.
In the opened, add an input property with the input you want to pass to your AgentKit network:Then, click on the “Run” button to trigger your AgentKit network”
Inspect AgentKit Agents token usage, input and output
In the run view of your AgentKit network run, the Agents step will be highlighted with a ✨ green icon.
By expanding the step, you can inspect the Agents:
The model used, ex: gpt-4o
The token usage detailed as prompt tokens, completion tokens, and total tokens
The input provided to the Agent
The output provided by the Agent
TipsYou can force line breaks to make the input and output more readable using the following button: You can expand the input and output view to show its full content using the following button: You can update the input of an AgentKit Agent and trigger a rerun from this step of the AgentKit network (see below)
On a given AgentKit Agent run, you can update the input of the Agent and trigger a rerun from this step of the AgentKit network.First, click on the “Rerun with new prompt” button under the input area.
Then, the following modal will open: