Agent toolingJot
New
Jot
A tool-running agent lets Jev choose actions and arguments from the current conversation.
Added to Jevfast
How it uses Jev
Jot gives Jev the conversation and available tools so Jev can choose the next tool and its arguments. Jot executes that call, adds the real result to context, and asks Jev again until Jev selects a reply. For example, the README shows Jev choosing calculator operations while the calculator performs the arithmetic.
What Jev decides
Decide whether the agent needs a longer replyExample answers · not a recorded Jev response · Source ↗
Question 1 · needs long draft
YOUR APP
INSTRUCTION
Decide whether answering the latest request needs a composed summary or explanation of tool or page evidence.
STATE
State contains decisionMessages from the conversation, available tools with their names and descriptions, and assistant identity. Successful prior tool results remain in the conversation. The illustrated Noul answer is not an observed response.
JEV · NOUL
YesNo
App workflow
- Prepare the agent context
The server provides the conversation and the available tool definitions to the Jev agent.
- Choose a tool call
Jev chooses a tool and its argument values from the agent's current state.
- Execute and record
Jot runs the selected tool and returns the actual tool result to the conversation context.
- Continue or reply
Jev is asked again with the updated context; the loop continues for further calls and ends when it chooses to answer.
Why it is interesting
It separates fast tool selection from tool execution and final prose. The same loop can chain actions using actual tool results, while the server owns credentials and the UI can expose the tool-call trace.