Skip to main content

​
Overview

Record what the user asked — their intent, questions, and requests — so future sessions have context about the user’s goals and communication patterns.
This tool is part of the agent profile and uses deferred loading to optimize performance.

​
Parameters

​
content
string
required
The user’s prompt textThis should be the raw user input — what they asked for, described, or requested. Future sessions can search prompts to understand user intent and preferences.
​
session_id
string
Session ID to associate with the promptDefault: manual-save-{project} or manual-save if no projectLinks the prompt to a specific coding session for chronological context.
​
project
string
Project nameAssociates the prompt with a specific project for filtering and context.

​
Response

​
result
string
Confirmation message with truncated prompt preview (first 80 characters)
Example: 
Prompt saved: "Add JWT authentication to the API and update the login endpoi..."

​
Usage Examples

​
Save User Request

{
  "content": "Add JWT authentication to the API and update the login endpoint to return tokens",
  "project": "my-api",
  "session_id": "session-2026-03-03-1430"
}
Response: 
Prompt saved: "Add JWT authentication to the API and update the login endpoint to return t..."

​
Save Without Session

{
  "content": "How does the search ranking algorithm work?",
  "project": "engram"
}
Response: 
Prompt saved: "How does the search ranking algorithm work?"

​
When to Use

  • At session start: Capture the initial user request to provide context
  • On major requests: When the user asks for something significant
  • For questions: When users ask about how things work
  • Goal tracking: Record what the user wanted to accomplish

​
What Gets Stored

Prompts are stored separately from observations and include:
  • Full prompt content (unlimited length)
  • Associated session ID
  • Project name
  • Timestamp
They can be searched via the HTTP API /api/prompts endpoint.

​
Privacy Considerations

Prompts capture raw user input. Be mindful of:
  • Sensitive information in user messages
  • API keys or credentials mentioned in prompts
  • Personal data in requests
Prompts follow the same project/personal scope as observations.

​
Why Save Prompts?

Future sessions benefit from knowing:
  • User intent: What were they trying to accomplish?
  • Communication style: How do they describe problems?
  • Context evolution: How did requirements change over time?
  • Repeated questions: What concepts need better documentation?

Build docs developers (and LLMs) love

Get started for freeTalk to us