Summary
Perplexity answers, in the window where the problem is.
The part that matters is the chat ID: an answer you got last week can be reopened and continued, and re-reading it costs nothing because the history lives on your machine rather than being fetched again.
What it is
A Python MCP server over the Perplexity API. It mirrors how you would use Perplexity in a browser — one-off questions, ongoing chats, and a list of everything you have asked — and keeps chat data locally.
What you get
- A one-off expert question answered, aimed at coding problems, error debugging and technical explanations, returned with source citations and alternative suggestions
- An ongoing conversation with full history context, returning a chat ID you can continue from later
- Every chat listed with IDs, titles and creation dates shown in relative time, a page at a time
- The complete history of one chat read back from local storage, with no API call made to Perplexity
- A different Perplexity model for asking than for chatting, if the two jobs deserve different ones
Requirements
Access to the Perplexity API. One environment variable picks the default model, and two more override it separately for asking and for chatting. Chat history is written to local storage on your machine, which is why re-reading a past conversation costs nothing.
Setup effort
One command plus a key — uvx mcp-perplexity, then supply credentials