VSCode-Perplexity-MCP brings Perplexity AI's search, reasoning, research, and compute capabilities into code editors through a Model Context Protocol (MCP) setup. Hosted at github.com/Automations-Project/VSCode-Perplexity-MCP, this TypeScript monorepo includes an MCP server, a dashboard, and auto-configuration tools for over 20 IDEs, with a dedicated extension for Visual Studio Code. It maintains a long-lived browser session with Perplexity AI, allowing developers to access these AI functions as native tools within their editing environment. The project holds 38 GitHub stars and lists on the MCP registry at registry.modelcontextprotocol.io. Note that it operates independently of Perplexity AI, Inc., and remains experimental—APIs and behavior may shift during development.

This setup addresses the gap between web-based AI tools like Perplexity and local development workflows. Instead of switching tabs or apps, users interact with Perplexity's modes (search, reason, research, compute) directly from their IDE. A demo video on YouTube shows this in action within VS Code, where queries trigger AI responses embedded in the editor.

Core components

The monorepo bundles several parts for a complete Perplexity MCP experience:

  • MCP server: Runs a persistent Perplexity browser session, exposing AI tools via the Model Context Protocol. Available as the npm package perplexity-user-mcp, with weekly downloads tracked on npmjs.com/package/perplexity-user-mcp.
  • VS Code extension: Published on the Visual Studio Marketplace as Nskha.perplexity-vscode, it integrates the MCP tools natively. Install counts appear on the marketplace badges.
  • Multi-IDE support: Auto-configuration scripts for 15+ IDEs and 20+ editors, including icons like VS Code's in the README table.
  • Dashboard: Provides a web interface for managing the session and tools.

Badges in the README highlight its status: CI passes on GitHub Actions, latest release available, recent commits on main, and MIT license. The project stays active, with the npm version badge showing updates.

Getting it running

Start with the VS Code extension, the primary entry point. Visit the marketplace at marketplace.visualstudio.com/items?itemName=Nskha.perplexity-vscode and click install. The badge provides a direct "Install in VS Code" link, which handles sideloading or marketplace addition.

For the backend MCP server:

  1. Install the npm package globally or locally:

    npm install -g perplexity-user-mcp

    Check the latest version on npmjs.com/package/perplexity-user-mcp.

  2. Run the server. The README implies a standard MCP server startup; consult the package docs for the exact command, typically something like npx perplexity-user-mcp or via a config file. It spins up the long-lived browser session.

  3. Configure your IDE. The monorepo includes auto-config for 20+ editors—run the setup scripts from the repo to point them at your local MCP server endpoint. For VS Code, the extension auto-detects or prompts for the server URL post-install.

The demo video demonstrates querying Perplexity modes from VS Code after setup, confirming tools like search and compute work inline. Ensure Node.js is installed, as it's TypeScript-based. Git clone the repo for full access:

git clone https://github.com/Automations-Project/VSCode-Perplexity-MCP
cd VSCode-Perplexity-MCP
npm install

From there, build and run packages like the extension or server via npm scripts (detailed in repo package.json files).

Troubleshoot via GitHub issues or the CI workflow at github.com/Automations-Project/VSCode-Perplexity-MCP/actions. Since it's experimental, test in a non-critical setup.

Who this fits

Developers using VS Code who rely on Perplexity AI for code-related queries benefit most. It suits tasks like researching APIs, reasoning through bugs, or computing solutions without leaving the editor. The multi-IDE auto-config appeals to teams with mixed setups—Cursor, Neovim, or others listed in the README icons.

Use cases include:

  • Inline AI search for library docs during coding.
  • Research mode for exploring unfamiliar frameworks.
  • Compute for quick calculations or data processing tied to your workspace.

If your workflow involves frequent Perplexity tabs, this embeds them via MCP, reducing context switches. Smaller projects or solo devs with 38-star tools like this find it lightweight compared to full AI suites.

How it stacks up

VSCode-Perplexity-MCP focuses narrowly on Perplexity via MCP, differing from broader AI extensions. GitHub Copilot or Cody offer code completion but lack Perplexity's search/reasoning depth. Other Perplexity integrations exist as browser extensions or CLI tools, but none match this monorepo's IDE auto-config and persistent session.

The npm package perplexity-user-mcp serves as the server backbone, similar to other MCP tools on registry.modelcontextprotocol.io, but tailored for Perplexity. It's lighter than self-hosted LLMs like Ollama integrations, relying on Perplexity's cloud with a local proxy. Drawbacks: experimental status means potential breaks on Perplexity updates; requires npm/Node.js, heavier than pure extensions.

For pure VS Code AI, marketplace alternatives like Continue.dev support multiple providers but need manual Perplexity API setup—no browser session persistence.

Practical limits and resources

This project works for hobbyists or early adopters testing AI in editors, but skip it for production pipelines due to the experimental warning. No enterprise guarantees, and changes could disrupt workflows. If stability matters, wait for maturity or use official Perplexity APIs directly.

Source code sits at github.com/Automations-Project/VSCode-Perplexity-MCP. Track releases, stars (currently 38), and npm stats. The YouTube demo at youtube.com/watch?v=wErgEe9Pgqo shows real usage.

View on GitHub →Official Website →