Web search

p0 can search the web when it needs external context — documentation, API references, error solutions, or information beyond its training data.

​

How it works

The AI has two web tools:

​

WebSearch

Returns a list of search results with titles and URLs:

• The AI constructs a search query based on what it needs
• Returns up to 10 results with titles, URLs, and snippets
• The AI can choose which results to fetch for detailed content

​

WebFetch

Retrieves full page content from specific URLs:

• Fetches the complete content of a web page
• Returns hostname, byte size, and HTTP status
• Content is converted to markdown for easier processing

The AI decides when to search and which results to fetch based on the task.

​

Result display

Search results appear as an expandable list:

• Expandable results: Click to expand and see all results
• Title and URL: Each result shows the page title and URL
• Link icon: Visual indicator that these are web resources
• Clickable URLs: Click any URL to open it in your browser

​

When AI searches

The AI typically performs web searches for:

• Unfamiliar APIs: Looking up API documentation for libraries it hasn’t seen before
• Error messages: Finding solutions to cryptic error messages or stack traces
• Library documentation: Checking the latest syntax or usage patterns for a framework
• Explicit requests: When you ask it to search for something specific

The AI uses its judgment about when external information would be helpful.

​

Privacy

Web searches are performed through Purple’s API with privacy protections:

• Search terms only: Only the search query is sent, never your code
• No code context: Your codebase content stays local and is not included in searches
• Secure fetching: WebFetch retrieves pages over HTTPS

For more details on how p0 protects your data, see the Security and Privacy documentation.