The focus modes are the lever: academicSearch and wolframAlphaSearch behave very differently from a plain web search, and choosing one is a parameter rather than a prompt trick. Nothing answers until Perplexica itself is running — this is the bridge, not the search engine.
A small Python server that talks to a Perplexica installation you run yourself. Two tools; every answer comes from your instance rather than a hosted service.
- An AI-powered search in a chosen focus mode — perplexica_search
- The chat and embedding models your instance has available — perplexica_get_models
- Six focus modes: webSearch, academicSearch, writingAssistant, wolframAlphaSearch, youtubeSearch and redditSearch
- Streaming responses, and conversation history carried across searches
A running Perplexica instance. Python 3.10 or newer; install with pip install git+https://github.com/armand0e/perplexica-mcp.git and run python -m perplexica_mcp. PERPLEXICA_BASE_URL points at the instance and defaults to http://localhost:3000.
One command — pip install git+https://github.com/armand0e/perplexica-mcp.git