Skip to main content

​
Overview

The AI chat interface provides a conversational way to interact with your vault and trigger automated actions. Powered by Claude, the chat agent understands natural language requests and routes them to the appropriate tool.

​
Available actions

ActionExample promptWhat happens
Password reset”Reset my Netflix password”Triggers the 3-phase reset orchestrator
Vault lookup”What’s my GitHub password?”Retrieves and decrypts the credential
Passkey lookup”Show my Google passkey”Retrieves passkey metadata
Share password”Share Netflix with sarah@email.com”Creates an encrypted share
Save credential”Save my new Figma login”Creates a new vault entry
Import passwords”Import from my CSV”Triggers the import flow
General chat”How strong is my security?”Provides security advice

​
How it works

1

Input guard

Your message is checked for injection attacks and sanitized before processing.
2

Intent detection

Claude Haiku with tool-use analyzes your message and determines the appropriate action.
3

Tool execution

The detected action is routed to the appropriate tool (vault lookup, reset orchestrator, sharing system, etc.).
4

Output guard

The response is checked before streaming to ensure no sensitive data leaks.
5

Streaming response

The response streams back as structured markdown sections that render as interactive UI components.

​
Streaming format

Responses use Server-Sent Events with structured markdown sections:
SectionRenders as
<!-- section:header -->Bold heading with context
<!-- section:progress -->Expandable agent accordion with step details
<!-- section:result -->Final success/failure message
<!-- section:captcha -->CAPTCHA escalation prompt
<!-- section:gmail -->Gmail search guidance
<!-- section:password -->Generated or reset password display

​
Conversation context

The chat maintains conversation history within a session. Follow-up messages understand context: 
You: "Reset my Instagram password"
Agent: [begins reset flow]
You: "The OTP code is 847293"
Agent: [submits OTP and continues]
You: "Done"
Agent: [confirms completion and saves new password]

​
Rate limits

  • 60 messages per 60 seconds per user
  • Reset tasks are additionally limited to 5 per 5 minutes
The AI chat is separate from the direct API endpoints. While the chat provides a conversational interface, all underlying operations use the same authenticated API with the same security controls.