perplexity-polza-mcp-server

# Perplexity MCP Server via Polza.ai `mcp-name: io.github.ivanantigravity-lgtm/perplexity-polza-mcp-server` MCP server for `Claude Desktop` and `Claude Code`, which interacts with Perplexity models (`Sonar`, `Sonar Pro`, `Sonar Reasoning`, `Sonar Deep Research`) through the aggregator [Polza.ai](https://polza.ai). ## Features - `perplexity_model_guide` — a cheat sheet for choosing a model for a task - `perplexity_ask` — a regular question in Perplexity - `perplexity_research` — more in-depth research with web search - `list_perplexity_models` — a list of available `perplexity/*` models from the Polza catalog ## Requirements for Installation - `Claude Desktop` or `Claude Code` (or any other MCP client) - [`uv`](https://docs.astral.sh/uv/) (installs with one command, see below) - Python 3.11+ - `POLZA_AI_API_KEY` — key taken from [polza.ai/dashboard/api-keys](https://polza.ai/dashboard/api-keys) Install `uv`: ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` ## Installation in 2 Minutes (via PyPI + uvx) This is the easiest way: no need to clone anything, `uvx` will download the package from PyPI. ### Claude Code / VS Code Create a file `.mcp.json` in the root of your project: ```json { "mcpServers": { "perplexity-polza": { "command": "uvx", "args": ["perplexity-polza-mcp-server@latest"], "env": { "POLZA_AI_API_KEY": "your-polza-api-key-here" } } } } ``` Restart Claude Code — done. ### Claude Desktop (macOS) Open the file `~/Library/Application Support/Claude/claude_desktop_config.json` and add: ```json { "mcpServers": { "perplexity-polza": { "command": "uvx", "args": ["perplexity-polza-mcp-server@latest"], "env": { "POLZA_AI_API_KEY": "your-polza-api-key-here" } } } } ``` Restart Claude Desktop. ### Claude Desktop (Windows) File: `%APPDATA%\Claude\claude_desktop_config.json`. The content is the same as on macOS. ## How to Check if it Works After restarting Claude, ask: > Show available Perplexity models through polza Claude should call the `list_perplexity_models` tool and return a list. ## Choosing a Model - `Sonar` — fast search + answer. News, facts, Q&A, short summaries. - `Sonar Pro` — denser structure, comparisons, follow-up questions. - `Sonar Pro Search` — deeper search, more search steps. - `Sonar Reasoning Pro` — not just find, but analyze and make a conclusion. - `Sonar Deep Research` — full-fledged research, market scan, long report. You can switch the default model using environment variables `PERPLEXITY_MODEL` and `PERPLEXITY_RESEARCH_MODEL` — the full list below. ## Environment Variables | Variable | Required | Default | Description | | --- | --- | --- | --- | | `POLZA_AI_API_KEY` | yes | — | Key from polza.ai | | `POLZA_BASE_URL` | no | `https://polza.ai/api/v1` | Base URL for chat completions (with `/v1` — it's expected, endpoint OpenAI-compatible) | | `PERPLEXITY_MODEL` | no | `perplexity/sonar` | Model for `perplexity_ask` | | `PERPLEXITY_RESEARCH_MODEL` | no | `perplexity/sonar-deep-research` | Model for `perplexity_research` | | `LOG_LEVEL` | no | `INFO` | — | ## Local Development ```bash git clone https://github.com/ivanantigravity-lgtm/perplexity-polza-mcp-server.git cd perplexity-polza-mcp-server uv sync POLZA_AI_API_KEY=your_key uv run python -m perplexity_polza_mcp_server.server ``` For local connection from source in Claude: ```json { "mcpServers": { "perplexity-polza-local": { "command": "uv", "args": ["run", "python", "-m", "perplexity_polza_mcp_server.server"], "cwd": "/absolute/path/to/perplexity-polza-mcp-server", "env": { "POLZA_AI_API_KEY": "your-polza-api-key-here" } } } } ``` ## Under the Hood - Chat completions: `POST https://polza.ai/api/v1/chat/completions` (OpenAI-compatible format) - Model catalog: `GET https://polza.ai/api/v1/models/catalog` ## Important Files in the Repository - `pyproject.toml` — package metadata and entry points - `server.json` — description for MCP Registry - `fastmcp.json` — local configuration for FastMCP - `.github/workflows/publish-pypi.yml` — publication in PyPI via GitHub Actions ## License MIT.