Web search, page fetch, AI extraction and deep research for AI agents, through the Sofya API.
Web search, page fetch, AI extraction and deep research for Dify Agents and Workflows, through the Sofya API.
The plugin is a thin client. Every tool sends one HTTPS request to and returns the response as a JSON message plus a readable text message.
The key is sent to Sofya as an header and nowhere else.
Outbound HTTPS to only. The plugin never connects to any other host, and the API endpoint is a constant in the code rather than a tool parameter. Request timeouts are 65 seconds for search, fetch and extract, and 190 seconds for research.
Searches the web and returns results that carry extracted page content, not only snippets.
Parameters: (required), ( default, or ), (1 to 20, default 10), (default false), , , ( default, or ), .
Example: , , . Returns 5 results with title, URL, published date and page content.
Fetches up to 10 URLs and returns each page as clean markdown. Web pages, PDF, DOCX and other document formats are supported. URLs that fail cost no credits.
Parameters: (required, comma separated, up to 10), (default false).
Example: . Returns the page as markdown.
Reads one page and extracts the information you ask for. If the page has no usable text, the answer comes back empty with set instead of an invented answer.
Parameters: (required), (required).
Example: , .
Splits a question into sub queries, reads many sources in parallel and returns a structured report with citations. A call can take a minute or two.
Parameters: (required), ( default, or ), , (5 to 30, default 20).
Example: , .
Current plans and credit prices are at https://sofya.co/pricing.
API errors are returned as a text message, for example an invalid key, an empty credit balance or a rate limit. The API key is never included in an error message.
MIT licensed.
Some outbound calls build their address at runtime, so the list may be incomplete.
Last checked 09/11/2026 09:19 AM · version 0.1.0