Skip to main content

Post Tool Call

In AI-driven workflows, the interaction between tools and the language model often requires contextual follow-up. Sometimes a tool succeeds, other times it fails — and you may want the LLM to react differently depending on the result.

To support this, CMND.ai introduces Post-Tool Call Prompting, a technique that allows you to dynamically guide the LLM’s response based on the outcome of a tool execution.

What It Does​

Each tool can define a postCallPrompt, which is injected into the system prompt after the tool completes but before the LLM responds to the user.

This gives you a chance to influence the assistant’s next message — based on success, failure, or specific outputs from the tool.

Why Use It?​

  • Adjust the assistant’s response based on real-world tool results
  • Handle success/failure cases with more clarity
  • Improve the flow of conversations without requiring custom post-processing

Example​

Let’s say you have a tool for handling user login. You want the assistant to prompt for retry on failure, or continue normally on success.

Tool with postCallPrompt
{
    "name": "user_login",
    "description": "Login tool that takes a username and password, and returns an access token",
    "parameters": custom_json_schema(UserLoginSchema),
    "runCmd": user_login,
    "isDangerous": False,
    "functionType": "backend",
    "isLongRunningTool": False,
    "postCallPrompt": "If the login tool output is failure, ask the user to try again. If successful, continue the conversation."
}

This prompt is merged with the system prompt and provides real-time context to the LLM, without needing any manual handling on your backend.

Best Practices
  • Be clear and specific in your postCallPrompt. The LLM needs precise instructions to behave consistently.
  • Use this feature for decision points in your flow — such as login, verification, search results, or validations.
  • Avoid duplicating logic that’s already enforced in your backend — use this for communication flow only.

Summary​

  • postCallPrompt lets you guide the LLM’s response after a tool call.
  • It supports smoother user experiences by reacting intelligently to tool outcomes.
  • Use it in any tool where the result should affect what the assistant says next.

This is especially powerful in assistant flows that involve real-time actions — such as authentication, transactions, or search — where context and control really matter.